A water and fertilizer integration optimization control method and system based on response surface analysis

Abstract

This invention discloses a water and fertilizer integration optimization control method and system based on response surface methodology. The method includes: using the encoded values ​​of single irrigation volume and irrigation liquid conductivity as independent variables, and the number of runners per mother plant and the number of daughter plants per mother plant as response variables, respectively, using the least squares method to independently fit two quadratic polynomial regression equations based on the same cultivation experiment data, thereby obtaining a first target response surface model and a second target response surface model; constructing a constraint region through the first target response surface model, and decoding the optimal encoded value combination feature corresponding to the predicted value of the maximum number of daughter plants within the constraint region using the second target response surface model, thereby obtaining the optimal control parameters for water and fertilizer integration, improving the accuracy of water and fertilizer integration control parameters, and adapting to the regulation needs of large-scale and standardized strawberry seedling propagation.

Description

Technical Field

[0001] This invention relates to the field of automated irrigation technology for agriculture, and relates to, but is not limited to, a water and fertilizer integration optimization control method and system based on response surface analysis. Background Technology

[0002] Strawberry seedling propagation via elevated substrate trays has become the mainstream production method for commercial strawberry seedlings due to its low risk of soil-borne diseases, high seedling uniformity, and ease of intensive operation. Water and fertilizer supply to the mother plant is crucial for controlling the number of runners and the efficiency of daughter plant differentiation. Insufficient water and fertilizer supply leads to sparse runner production and an inability to meet the nutrient needs of the mother plant, while excessive water and fertilizer supply causes excessive growth of the mother plant, inhibiting daughter plant differentiation and causing salt stress. Therefore, precise control of the water supply to the mother plant is a key factor determining the efficiency of strawberry seedling propagation.

[0003] Currently, the regulation of water supply to strawberry mother plants mainly relies on the grower's experience. This experience-based management method cannot quantify the differential coupling effect of irrigation volume and nutrient solution concentration on the two reproductive stages of stolon emergence and seedling differentiation. Furthermore, the parameter setting is highly subjective and has poor reproducibility, resulting in large fluctuations in seedling quality between batches. It is impossible to maximize seedling production while ensuring the minimum number of reproductive channels, and it is difficult to adapt to the precise regulation requirements of large-scale and standardized seedling production. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a water and fertilizer integration optimization control method based on response surface analysis, aiming to overcome the defects of the prior art and provide a highly adaptable and quantifiable water and fertilizer integration optimization control method based on response surface analysis to meet the regulation needs of large-scale and standardized seedling breeding of strawberry seedlings.

[0005] The specific technical solutions of this invention are as follows: This invention provides a water and fertilizer integration optimization control method based on response surface methodology, comprising: The single irrigation volume and the conductivity of the irrigation solution to be processed are obtained to form data to be processed, and the data to be processed is subjected to dimensionless encoding conversion to generate combined features to be processed. Based on the characteristics of the combination to be processed, the predicted value of the number of stolons in the value space of the control factors is calculated using the first target response surface model. The constrained region is constructed by combining the predicted value of the number of stolons, the target reproductive channel threshold, and the characteristics of the combination to be processed. Within the constrained region, the extreme value solution of the second target response surface model is performed to obtain the optimal combination of encoded values ​​that maximizes the predicted number of seedlings; The optimal combination of encoded values ​​is decoded to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity. The optimal single irrigation volume and the optimal irrigation liquid conductivity are used as the optimal control parameters for fertigation.

[0006] This invention provides a water and fertilizer integration optimization control system based on response surface analysis. The system includes a combined feature generation module, a constraint region construction module, an optimal solution solving module, and a feature decoding module, wherein: The combined feature generation module is used to obtain the single irrigation volume to be processed and the conductivity of the irrigation liquid to be processed to form the data to be processed, and to perform dimensionless encoding conversion processing on the data to be processed to generate the combined features to be processed. The constraint region construction module is used to calculate the predicted number of stolons in the control factor value space based on the features of the combination to be processed using the first target response surface model, and to construct the constraint region by combining the predicted number of stolons, the target reproductive channel threshold and the features of the combination to be processed. The optimal solution solving module is used to solve the extreme value of the second target response surface model within the constrained region to obtain the optimal combination of encoded values ​​that maximizes the predicted value of the number of seedlings. The feature decoding module is used to decode the optimal combination of encoded values ​​to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity, and to use the optimal single irrigation volume and the optimal irrigation liquid conductivity as the optimal control parameters for fertigation.

[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, firstly, the single irrigation volume and the conductivity of the irrigation solution are selected as control factors. Then, a two-factor quadratic general rotational combination design is used to generate multiple factor level combination schemes. Next, using the encoded values ​​of the single irrigation volume and the encoded values ​​of the irrigation solution conductivity as independent variables, and the number of runners per mother plant and the number of daughter plants per mother plant as response variables, based on the same cultivation experiment data, two quadratic polynomial regression equations are independently fitted using the least squares method to obtain a first target response surface model and a second target response surface model. A constraint region is constructed using the first target response surface model, and within the constraint region, the optimal encoded value combination feature corresponding to the predicted value of the maximum number of daughter plants is obtained using the second target response surface model. This is then decoded to obtain the optimal control parameters for integrated water and fertilizer management, improving the accuracy of the integrated water and fertilizer management parameters to meet the regulatory needs of large-scale and standardized strawberry seedling propagation. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 A schematic flowchart of the water and fertilizer integration optimization control method based on response surface analysis provided in an embodiment of the present invention; Figure 2 The surface plot and contour plot of the first target response surface model provided in the embodiments of the present invention; Figure 3 The surface plot and contour plot of the second target response surface model provided in the embodiments of the present invention; Figure 4 This is a schematic diagram illustrating the optimal combination feature optimization provided in an embodiment of the present invention. Figure 5 This is a system block diagram of a water and fertilizer integration optimization control system based on response surface analysis, provided as an embodiment of the present invention. Detailed Implementation

[0009] 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. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention. 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.

[0010] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0011] It should be noted that the terms "first, second, and third" used in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.

[0012] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments of the invention pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0013] This invention relates to an optimal control method for integrated water and fertilizer management based on response surface methodology, applied to the elevated substrate cultivation of strawberries. In this method, mother seedlings produce daughter seedlings by producing runners, which are then cut and propagated in seedling trays for production. The water and fertilizer supply to the mother seedlings directly affects the number of runners and the number of daughter seedlings. In this embodiment, the water and fertilizer supply factors include the amount of water per irrigation and the conductivity of the irrigation solution. The amount of water per irrigation directly affects the moisture content of the cultivation substrate, thus determining the efficiency of water and nutrient absorption by the mother seedling roots. The conductivity of the irrigation solution, as a proxy variable for the overall nutrient level of the irrigation nutrient solution, reflects the total concentration of soluble salt ions in the nutrient solution. Insufficient water per irrigation leads to sparse runner production, while excessive water per irrigation results in excessive vegetative growth and inhibits daughter seedling differentiation. Too low a concentration of the irrigation nutrient solution cannot meet the nutrient requirements of the mother seedlings, while too high a concentration causes salt stress. Therefore, precise control of irrigation water volume and nutrient solution concentration is a key factor in determining the efficiency of strawberry seedling propagation.

[0014] Since the total nutrient concentration of irrigation nutrient solution cannot be directly detected by online sensors, and the concentration of individual nutrient elements requires laboratory testing, there is a significant detection lag, which cannot support real-time closed-loop control of automated irrigation. Therefore, this embodiment uses irrigation solution conductivity as a quantitative proxy indicator of the comprehensive nutrient level of the irrigation nutrient solution, based on the stable positive correlation between irrigation solution conductivity and total nutrient concentration. Specifically, irrigation solution conductivity can be detected in real-time and continuously using a standardized conductivity sensor, which can directly interface with the standard control interface of existing integrated water and fertilizer equipment. This eliminates the need for separate detection components for individual elements such as nitrogen, phosphorus, and potassium in the irrigation nutrient solution, making it more suitable for elevated substrate soilless cultivation scenarios.

[0015] like Figure 1 The diagram shown is a flowchart of a water and fertilizer integration optimization control method based on response surface methodology provided in an embodiment of the present invention. The method includes at least the following steps: Step S110: Obtain the single irrigation volume and the conductivity of the irrigation solution to be processed to form the data to be processed, and perform dimensionless encoding conversion on the data to be processed to generate the combined features to be processed.

[0016] Among them, the single irrigation volume to be processed refers to the single round of irrigation water volume required in the process of determining the optimal control parameters of water and fertilizer integration, and the value range is 229 mL / time to 771 mL / time, so as to ensure that the model prediction is within the effective fitting range.

[0017] The conductivity of the irrigation solution to be treated refers to the conductivity of the irrigation solution that needs to be used in determining the optimal control parameters for fertigation. The value range is 500 μs / cm to 1900 μs / cm to ensure that the model prediction is within the effective fitting range.

[0018] Specifically, the range of values ​​for the single irrigation volume to be processed is first set to... The range of values ​​for the conductivity of the irrigation solution to be treated is set as follows: Then, gridded full-traversal sampling is performed according to a fixed step size to generate data to be processed. After obtaining the data to be processed, in order to achieve unified processing of two different dimensional factors, namely the single irrigation volume and the conductivity of the irrigation liquid, in the same mathematical model, this embodiment also needs to encode and convert the above two factors to generate the combined features to be processed.

[0019] The specific process of encoding conversion is as follows: The core baseline parameters after encoding conversion of the single irrigation volume and the conductivity of the irrigation solution to be processed are calculated separately, including the central physical quantity corresponding to the encoding zero point and the unit encoding step size. Then, based on the central physical quantity and the unit encoding step size, a linear mapping conversion is performed on each set of physical quantities (single irrigation volume and irrigation solution conductivity) in the data to be processed to obtain the combined features to be processed. Within the range of values ​​for single irrigation volume and irrigation solution conductivity, all parameters follow the two-factor quadratic general rotational combination design specification to ensure complete consistency with the encoding system in the response surface modeling stage.

[0020] The formula for calculating the central physical quantity is as follows:

[0021] In the formula, Factor coding index indicating the amount of irrigation water to be treated and the conductivity of the irrigation solution to be treated. The time indicates the amount of water to be processed in a single irrigation cycle. The time indicates the conductivity of the irrigation solution to be treated; Indicates the first The central physical quantity of the factor-coded index Indicates the first The lower bound of the coding value boundary corresponding to the factor coding index. Indicates the first The upper limit of the encoding value boundary corresponding to each factor encoding index.

[0022] The formula for calculating the unit coding step size is:

[0023] In the formula, Factor coding index indicating the amount of irrigation water to be treated and the conductivity of the irrigation solution to be treated. The time indicates the amount of water to be processed in a single irrigation cycle. The time indicates the conductivity of the irrigation solution to be treated; Indicates the first The unit coding step size corresponding to the factor coding index; , is the standard asterisk arm length for a two-factor, two-dimensional universal rotational combination design.

[0024] The formula for linear mapping transformation is:

[0025] In the formula, Indicates the first The coded value of a single irrigation volume to be processed. Indicates the first The coded value of the conductivity of the irrigation solution to be treated. Indicates the first One volume of irrigation water to be processed per cycle Indicates the first The conductivity of the irrigation solution to be treated This represents the central physical quantity corresponding to the amount of irrigation water to be processed in a single instance. This represents the central physical quantity corresponding to the conductivity of the irrigation solution to be treated. This indicates the unit coding step size corresponding to the single irrigation volume to be processed. This indicates the unit coding step size corresponding to the conductivity of the irrigation solution to be treated.

[0026] After the encoding conversion process is completed, the combined features to be processed are represented as follows: .

[0027] Furthermore, in this embodiment, the unit encoding step size is determined by the difference between the ±1 level actual physical quantity and the central physical quantity. Here, the ±1 level actual physical quantity refers to the actual parameter value with physical units corresponding to the encoding value of +1 and -1 (including the single irrigation volume to be treated and the conductivity of the irrigation solution to be treated).

[0028] For ease of understanding, this embodiment uses the data in Tables 1 and 2 to illustrate the single irrigation volume to be treated, the conductivity of the irrigation solution to be treated, and their respective coded values. It should be noted that the specific values ​​in Tables 1 and 2 are merely illustrative and are not intended to limit the scope of protection of this application's embodiments.

[0029] Table 1

[0030] Table 2

[0031] The unit coding step size for the single irrigation volume to be processed is calculated based on the data in Table 1: =500-350=150mL / time. The unit coding step size of the conductivity of the irrigation solution to be treated is calculated using the data in Table 2: Δ=1200-700=500μs / cm.

[0032] The asterisk point is horizontal (±) The values ​​are set directly based on the actual tolerance limit of the seedlings and the controllable range of the irrigation equipment. According to the data in Table 1, the star points corresponding to the single irrigation volume to be treated are 229 mL / time and 771 mL / time, and the star points corresponding to the conductivity of the irrigation liquid to be treated are 500 μs / cm and 1900 μs / cm.

[0033] This design retains the experimental point structure of a two-factor quadratic universal rotational combination design, including: factorial points (coded values ​​of ±). (dots), asterisks (coded values ​​±) The points (points with a coded value of 0) and center points ensure the rotation and orthogonality of the subsequent response surface model, while making the physical quantities at each level more consistent with the operable accuracy of the equipment in actual production.

[0034] Step S120: Based on the features of the combination to be processed, the predicted value of the number of stolons in the value space of the control factors is calculated using the first target response surface model. The constrained region is constructed by combining the predicted value of the number of stolons, the target reproductive channel threshold and the features of the combination to be processed.

[0035] Among them, the control factor value space refers to the value range of a single irrigation volume and the value range of the irrigation liquid conductivity, which are set according to the actual situation.

[0036] In one embodiment, the construction process of the first target response surface model is as follows: 1) The single irrigation volume and the conductivity of the irrigation solution were selected as the modeling control factors, and the water and fertilizer treatment parameters were determined by the point layout rule of the two-factor quadratic general rotational combination design.

[0037] Specifically, the modeling single irrigation volume and the modeling irrigation fluid conductivity are selected as two modeling control factors. The central physical quantity of the modeling control factor is determined according to the calculation formula of the central physical quantity (the formula form is as shown in step S110, and will not be repeated here). The unit coding step size of the modeling control factor is determined according to the calculation formula of the unit coding step size (the formula form is as shown in step S110, and will not be repeated here), and is represented by the standard asterisk arm length. As the boundary for encoding values, the distribution rules of the two-factor quadratic general rotational combination design are used to generate the encoding combination vector formed by the factorial point, the asterisk point and the center point. Finally, the above encoding combination vector is converted into physical quantities (that is, the encoded value is mapped to the actual water and fertilizer parameters) through the linear encoding conversion rule to obtain the water and fertilizer treatment parameters.

[0038] For ease of understanding, this embodiment uses the data in Table 3 to illustrate the factorial points, asterisk points, and center points, as well as the water and fertilizer parameters obtained by mapping each point. It should be noted that the specific values ​​in Table 3 are for illustrative purposes only and are not intended to limit the scope of protection of this application embodiment.

[0039] Table 3

[0040] In Table 3, numbers 1-4 are factorial points, numbers 5-8 are asterisk points, and numbers 9-13 are center points.

[0041] 2) Obtain the measured data of the number of stolons per mother plant obtained from the cultivation experiment based on water and fertilizer treatment parameters, and generate the first experimental sample set by combining the modeling control factors.

[0042] Specifically, after determining the water and fertilizer treatment parameters, in order to reduce the impact of individual differences and random errors on the results, except for the modeling of single irrigation volume and modeling of irrigation liquid conductivity, other management measures (such as light, temperature, and pest and disease control) were kept strictly consistent in the cultivation experiment to ensure that the differences in the experimental results were caused only by the modeling of single irrigation volume and modeling of irrigation liquid conductivity.

[0043] In this embodiment, the strawberry cultivation method is elevated substrate cultivation, and the cultivation substrate is a pre-prepared formula of coconut coir, peat moss, vermiculite, and perlite. After determining the modeling single irrigation volume and the modeling irrigation solution conductivity, a cultivation experiment was carried out based on the modeling single irrigation volume and modeling irrigation solution conductivity. After the period of abundant runner growth (generally 60-90 days after the mother seedlings were planted), the number of runners per mother seedling in each group was manually counted. The number of runners produced by each mother seedling that reached the standard for cutting daughter seedlings (≥10cm) was recorded as the measured data of the number of runners per mother seedling. Finally, the first experimental sample set was formed based on the modeling single irrigation volume, modeling irrigation solution conductivity, and the measured data of the number of runners per mother seedling.

[0044] 3) The least squares method is used to fit the quadratic polynomial regression relationship of the first experimental sample set to construct the first original response surface model; Specifically, after determining the first experimental sample set, the modeling single irrigation volume code value and the modeling irrigation liquid conductivity code value are used as independent variables, and the number of stolons produced by a single mother seedling (i.e. the number of stolons per single mother seedling) is used as the response variable. The least squares method is used to fit a quadratic polynomial regression equation (factor analysis points estimate the coefficients of the first term and interaction term of the response surface model, asterisk points estimate the coefficients of the quadratic term of the response model, and the center point is used to estimate the model error) to construct the first original response model.

[0045] 4) Perform a lack of fit test and regression significance verification on the first original response surface model. If both the lack of fit test and regression significance verification pass, then the first original response surface model is taken as the first target response surface model.

[0046] Specifically, after constructing the first original response surface model, the F-values ​​of the lack-fit term and regression F-values ​​of the first original response surface model are calculated. A lack-fit test is performed based on the calculated F-values. If the F-value is less than a preset lack-fit threshold, the lack of fit is not significant, the lack-fit test passes, meaning the unfitted portion of the model does not constitute significant variation, and the model form is reasonably chosen. A regression significance verification is performed based on the calculated F-values. If the F-value is greater than a preset regression threshold, the regression relationship is significant, the regression significance verification passes, meaning the independent variables as a whole can significantly explain the variation in the response variable, and the model is reliable.

[0047] The formula for calculating the F-value of the lack-of-fit term is as follows:

[0048] In the formula, express ; This indicates the number of stolons per single mother plant. This is used to reflect the degree of variation caused by the mismatch between the functional form of the model and the actual response pattern; This indicates the number of stolons per single mother plant. It is used to reflect the degree of pure random fluctuation variation caused by repeated experiments.

[0049] The formula for calculating the regression F-value is:

[0050] In the formula, express ; This indicates the number of stolons per single mother plant. This is used to reflect the total variability of the response variables that the model can explain; This indicates the number of stolons per single mother plant. This is used to reflect the degree of total variation in the residuals of the model.

[0051] If both the above-mentioned failure test and regression significance verification are passed, then the first original response surface model is taken as the first target response surface model.

[0052] The first objective response surface model is:

[0053] In the formula, This represents the predicted number of runners per individual mother plant. This represents the constant intercept term of the first objective response surface model. This represents the linear effect coefficient of a single change in irrigation volume on the number of runners per mother plant. This represents the dimensionless coded value corresponding to the single irrigation volume to be processed. This represents the linear effect coefficient of the change in the conductivity of the irrigation solution on the number of runners per mother plant. This represents the dimensionless coded value corresponding to the conductivity of the irrigation solution to be treated. This represents the surface effect coefficient that affects the number of runners per mother plant when the amount of irrigation changes in a single instance. The surface effect coefficient represents the effect of changes in the conductivity of the irrigation solution on the number of runners per mother plant. This represents the interaction term regression coefficient that affects the number of stolons per mother plant when the amount of water used in a single irrigation and the conductivity of the irrigation solution change in tandem.

[0054] It should be noted that, and The coefficient of the linear term, and The coefficient of the quadratic term, For the interaction term coefficient.

[0055] The first target response surface model described above will be illustrated below with a specific embodiment. However, it is worth noting that this specific embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.

[0056] pass The fitted first original response surface model is as follows:

[0057] The F-value for the lack of fit of the first original response surface model was calculated to be 0.095, and the F-value for regression was 4.006. The F-value of 0.095 is less than the preset lack of fit threshold of 6.59, indicating that the lack of fit is not significant and the lack of fit test is passed. This means that the unfitted portion of the model does not constitute significant variation, the model form is reasonably chosen, and it can reflect the true functional relationship. The F-value for regression of 4.006 is greater than the preset regression threshold of 3.97, indicating that the regression relationship is... If the value is significant at the 0.10 level, and the regression significance verification is passed, then this first original response surface model is used as the first target response surface model. For example... Figure 2 As shown, the left figure is the surface plot of the first target response surface model (Response Surface of Stolon number (Y1)), and the right figure is the contour plot of Y1. (The figures are incomplete.) This represents the predicted number of stolons per mother plant. This represents the dimensionless code value (water code) corresponding to the single irrigation volume to be processed. This represents the dimensionless code value (EC code) corresponding to the electrical conductivity of the irrigation solution to be treated. EC stands for Electrical Conductivity.

[0058] The pentagram in the right-hand diagram represents the unconstrained optimal point: That is, within the range of experimental parameters, the maximum number of stolons per mother plant, 10.73, appeared at the center point of the coding. =0, =0), corresponding to a single irrigation volume of 500mL / time and an irrigation solution conductivity of 1200μs / cm.

[0059] In one embodiment, the propagation process of strawberry mother seedlings has a clear stage logic: the mother seedling first produces runners (intermediate propagation), and then the runners segment to produce daughter seedlings (the final propagation output). Runners are the physical carriers for daughter seedling production. If the number of runners is insufficient, even if the daughter seedling differentiation conditions on each runner are suitable, the total number of daughter seedlings will be structurally limited due to insufficient channels. Based on this, after constructing the first target response surface model, this embodiment of the application also needs to determine that the predicted number of runners is not lower than the target propagation channel threshold. This target propagation channel threshold can be set according to the characteristics of the strawberry variety and the economic requirements of intensive seedling cultivation; for example, it can be set to 8 runners / plant to ensure that the propagation coefficient of a single plant meets certain economic requirements.

[0060] In step S120, combining the predicted number of stolons, the target reproductive channel threshold, and the characteristics of the combination to be processed, a constraint region is constructed using the laws of reproductive biology. This specifically includes the following steps: 1) Determine the target propagation channel threshold based on the planting density of mother seedlings, the target number of daughter seedlings per mu, and the number of daughter seedlings per single stolon.

[0061] Among them, the target propagation channel threshold refers to the minimum number of stolons set to ensure the differentiation requirements of seedlings.

[0062] Specifically, the formula for calculating the target reproductive channel threshold is as follows:

[0063] In the formula, Indicates the target reproductive channel threshold. This indicates the target yield of seedlings per acre. Indicates the planting density of mother seedlings. This indicates the number of seedlings on a single creeping stem.

[0064] Among them, the target yield of seedlings per mu The specific parameters can be set by the grower based on the characteristics of the strawberry variety and production goals. According to the project's cultivation trial data, when a single mother plant produces 10.73 runners and 39.25 daughter plants, which translates to an average of approximately 3.66 cuttings per runner, =4000 plants / 667 , =117000 plants / 667 When r = 3.66, T ≈ 8 plants / plant. If a higher density is used (e.g., 5000 plants / 667...),... If a lower density is used (e.g., 3000 plants / 667), then T can be adjusted accordingly to approximately 6.5 lines / plant; if an even lower density is used (e.g., 3000 plants / 667), T can be adjusted accordingly to approximately 6.5 lines / plant; If the target number of runners is less than 10.7, then T needs to be adjusted to approximately 10.7 per plant. The target propagation channel threshold, determined by the planting density of the mother plant, the target number of daughter plantlets per acre, and the number of daughter plantlets per single runner, can adaptively adjust with changes in planting density and target yield. Therefore, assuming an average of 3-4 cuttings per runner, 8 runners correspond to approximately 24-32 daughter plantlets, resulting in a propagation coefficient of approximately 1:24 to 1:32 per plant. If the number of runners is less than 8, the overall propagation coefficient will be insufficient to meet the economic requirements of intensive seedling cultivation (typically requiring a ratio of 1:25 or higher).

[0065] 2) Obtain the predicted number of stolons corresponding to all the combined features to be processed, as well as the coded value boundaries corresponding to the single irrigation volume and the conductivity of the irrigation solution to be processed. Furthermore, a constraint screening criterion is generated by combining the target reproductive channel threshold T.

[0066] 3) Based on the constraint screening and judgment criteria, condition matching is performed on each group of features to be processed to extract the target combination feature vectors where the predicted number of stolons is not lower than the target reproductive channel threshold and the corresponding encoded value is within the encoding value boundary.

[0067] 4) Summarize and integrate all extracted target combination feature vectors to obtain the constrained region. .

[0068]

[0069] In the formula, Indicates the constraint region. Let represent the target combined feature vector, where The coded value representing the amount of irrigation water to be processed in a single instance. The coded value representing the conductivity of the irrigation solution to be treated; This is a set syntax separator, indicating that all of the following conditions must be met; This indicates the coded value based on the single irrigation volume to be processed. and the conductivity of the irrigation solution to be treated The calculated predicted number of stolons, Indicates the target reproductive channel threshold, [ [] indicates the boundary of the encoded value.

[0070] Step S130: Within the constraint region, the extreme value solution of the second target response surface model is performed to obtain the optimal combination of coding values ​​that maximizes the predicted number of seedlings.

[0071] In one embodiment, the construction process of the second target response surface model is as follows: 1) The single irrigation volume and the conductivity of the irrigation solution were selected as the modeling control factors, and the water and fertilizer treatment parameters were determined by the point layout rule of the two-factor quadratic general rotational combination design.

[0072] Specifically, based on the same modeling single irrigation volume and modeling irrigation liquid conductivity as in step S120, the water and fertilizer treatment parameters are determined using the same two-factor quadratic general rotational combination design point placement rules as in step S120. For the specific determination and execution process, please refer to the corresponding steps of the first target response surface model. To avoid repetition, it will not be described again here.

[0073] 2) Obtain the measured data of the number of daughter seedlings produced by a single mother seedling during the cultivation experiment based on water and fertilizer treatment parameters, and generate a second experimental sample set by combining the modeling control factors.

[0074] Specifically, after determining the water and fertilizer treatment parameters, a cultivation experiment identical to step S120 is conducted based on the water and fertilizer treatment parameters. The measured data of the number of daughter seedlings produced by a single mother seedling are collected, and a second experimental sample set is generated by combining the modeling control factors.

[0075] 3) The least squares method was used to fit the second experimental sample set with a quadratic polynomial regression relationship to construct the second original response surface model.

[0076] Specifically, after determining the second experimental sample set, the modeling single irrigation volume code value and the modeling irrigation liquid conductivity code value are used as independent variables, and the number of daughter seedlings produced by a single mother seedling is used as the response variable. The second original response model is constructed by fitting a quadratic polynomial regression equation using the same least squares method as in step S120 (factor analysis points estimate the coefficients of the first term and interaction term of the response surface model, asterisk points estimate the coefficients of the quadratic term of the response model, and center points are used to estimate the model error).

[0077] 4) Perform a lack of fit test and regression significance verification on the second original response surface model. If both the lack of fit test and regression significance verification pass, then the second original response surface model is used as the second target response surface model.

[0078] Specifically, after constructing the second original response surface model, the F-values ​​of the lack-of-fit term and regression F-values ​​of the second original response surface model are calculated. A lack-of-fit test is performed based on the calculated F-values. If the F-value is less than a preset lack-of-fit threshold, it indicates that the lack of fit is not significant, the lack-of-fit test passes, and the model form is reasonably chosen. A regression significance verification is performed based on the calculated F-values. If the F-value is greater than a preset regression threshold, it indicates that the regression relationship is significant, the regression significance verification passes, and the model is reliable.

[0079] The calculation of the F-value of the missing term and the F-value of the regression in the second original response surface model refers to the calculation formula of the F-value of the missing term and the calculation formula of the regression F-value in step S120. The only difference is that the number of stolons per mother plant in the two formulas is replaced with the number of daughter plants produced by a single mother plant.

[0080] The second objective response surface model is:

[0081] In the formula, This represents the predicted number of daughter plantlets produced by a single mother plantlet. This represents the constant intercept term of the first objective response surface model. This represents the linear effect coefficient of a single change in irrigation volume on the number of daughter seedlings produced by a single mother seedling. This represents the dimensionless coded value corresponding to the single irrigation volume to be processed. This represents the linear effect coefficient of the change in the conductivity of the irrigation solution on the number of daughter seedlings produced by a single mother seedling. This represents the dimensionless coded value corresponding to the conductivity of the irrigation solution to be treated. This represents the surface effect coefficient on the number of daughter seedlings produced by a single mother seedling when the amount of irrigation changes. This represents the surface effect coefficient on the number of daughter seedlings produced by a single mother seedling when the conductivity of the irrigation solution changes. This represents the regression coefficient of the interaction term generated by the single mother seedling when the amount of water used for irrigation changes in tandem with the conductivity of the irrigation solution.

[0082] The second objective response surface model described above will be illustrated below with a specific embodiment. However, it is worth noting that this specific embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.

[0083] pass The fitted second original response surface model is as follows:

[0084] The F-value for the lack of fit of the second original response surface model was calculated to be 0.043, and the F-value for regression was 4.51. The F-value of 0.043 is less than the preset lack of fit threshold of 6.59, indicating that the lack of fit is not significant and the lack of fit test is passed. This means that the unfitted portion of the model does not constitute significant variation, the model form is reasonably chosen, and it can reflect the true functional relationship. The F-value for regression of 4.51 is greater than the preset regression threshold of 3.97, indicating that the regression relationship is... If the value is significant at the 0.10 level, and the regression significance verification is passed, then this second original response surface model is used as the second target response surface model. For example... Figure 3 As shown, the left figure is the response surface of the second objective response surface model (Y2), and the right figure is the contour plot of Y2. This represents the predicted number of seedlings produced by a single mother plant. This represents the dimensionless code value (water code) corresponding to the single irrigation volume to be processed. This represents the dimensionless code value (EC code) corresponding to the electrical conductivity of the irrigation solution to be treated. EC stands for Electrical Conductivity.

[0085] The pentagram in the right-hand diagram represents the unconstrained optimal point: That is, within the range of experimental parameters, the maximum predicted number of daughter seedlings produced by a single mother seedling, 39.25 per seedling, appeared at the coding center point. =0, =0), corresponding to a single irrigation volume of 500mL / time and an irrigation solution conductivity of 1200μs / cm.

[0086] Specifically, after determining the constraint region, the function value of the second objective response surface model is searched for the maximum value within the constraint region to determine the dimensionless encoded values ​​of the single irrigation volume and irrigation liquid conductivity corresponding to the maximum predicted value of the number of daughter seedlings produced by a single mother seedling, which are used as the optimal encoded value combination features.

[0087] Furthermore, since the second objective response surface model is a smooth, continuous quadratic polynomial function, and the constrained region is usually a convex set, the stationary points where the partial derivatives are zero can be directly calculated using the quadratic function extremum solution method. Determine the optimal combination of encoded values. If the stationary point is within the constraint region (i.e., And the stationary points of the second objective response surface model ( ) at the boundary of encoded values ​​[ If the stationary point is not within the constraint region, then the value corresponding to the stationary point is directly used as the optimal encoding value combination feature; if the stationary point is not within the constraint region, that is, the predicted value of the number of stolons of a single mother plant corresponding to the unconstrained objective solution. If the value is below the target reproductive channel threshold T, then the target solution lies on the constraint boundary (i.e., ...). At this point, a one-dimensional search can be performed along the constraint boundary to find the predicted number of seedlings. The largest boundary point is used as the optimal combination of encoded values. This hierarchical constraint optimization architecture ensures that the number of daughter plants produced by a single mother plant is maximized while meeting the basic requirements of the propagation channel.

[0088] The one-dimensional search process in this embodiment is as follows: 1) Determine the one-dimensional search path like This indicates that the station is located at To the right of the boundary, it should be along The boundary, for Perform a univariate incremental search, that is, with To scan variables, search upwards along fixed boundaries; like This indicates that the station is located at On the left side of the boundary, it should be along The boundary, for Perform a univariate incremental search, that is, with To scan variables, search upwards along fixed boundaries; like This indicates that the station is located at Above the boundary, it should be along The boundary, for Perform a univariate decreasing search, i.e. To scan variables, search downwards along fixed boundaries; like This indicates that the station is located at Below the boundary, it should be along The boundary, for Perform a univariate incremental search, that is, with To scan variables, search upwards along fixed boundaries.

[0089] 2) Construction of one-dimensional search path Select the point closest to the stationary point as the starting point of the search according to the search step size. ), from the search starting point ( Starting with a fixed variable, the search points are determined by performing univariate transformations on another variable according to the search step size. For each search point... calculate And determine whether it meets the requirements. If the conditions are met, then calculate the corresponding predicted number of seedlings. And record all that meet the requirements. The scan points, among which, are selected to make The point that is maximized is taken as the optimal boundary point, and its corresponding value is the optimal combination feature of encoded values.

[0090] Furthermore, if multiple scan points have the same maximum If the value is less than or equal to the stationary point, then the point closer to the stationary point is selected as the optimal boundary point to ensure that the optimal boundary point is the globally optimal boundary point.

[0091] The calculation process of the above-mentioned optimal coding value combination feature is described below with reference to a specific embodiment. However, it should be noted that this specific embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.

[0092] When the target reproductive channel threshold T is 8, in Within the constraint region, for the second response surface model Perform a maximum value search. Based on the specific numerical value... As can be seen from the properties of the quadratic polynomial, It reaches its maximum value at the stationary point (the point where both partial derivatives are zero). Calculate the partial derivatives:

[0093]

[0094] Solving the system of equations yields a stationary point of approximately (0.14, 0.20), located near the center point. Further calculations are needed to determine the corresponding... The value is 39.53, which is the value at the center point (0, 0). The value 39.25 is extremely close (the difference is less than 0.7%). Considering the simplicity of the engineering implementation, the center point (0, 0) is taken as the optimal solution. This point simultaneously satisfies... The constraints on Y1 mean that it lies within the constraint region. Therefore, the optimal combination of encoded values ​​is... , . Figure 4 For the optimal combination of encoded values ​​features , The diagram illustrates constraint optimization. Among them, This represents the dimensionless code value (water code) corresponding to the single irrigation volume to be processed. This represents the dimensionless code value (EC code) corresponding to the electrical conductivity of the irrigation solution to be treated. EC stands for Electrical Conductivity.

[0095] In one embodiment, a water and fertilizer integration optimization control method based on response surface analysis further includes a failure handling process for misfit test; The process for handling failures in the lack of similarity test is as follows: Step 1: If the lack-of-fit test fails, confirm whether there is outlier data in the experimental sample set; if outlier data exists, remove the outlier data and obtain the experimental sample set after performing the outlier operation as the corrected sample set; in some implementations, the Grubbs test is used to confirm whether there is outlier data in the experimental sample set; given a significance level. Calculate the residuals of each sample point in the experimental sample set. If the absolute value of the residual is greater than the product of the Grubbs critical value and the standard deviation of the residual, it is judged as outlier data and removed. Step 2: Refit the quadratic polynomial regression relationship based on the corrected sample set to construct a new response surface model; Step 3: If there are no outliers in the experimental sample set, add higher-order terms (i.e., add higher-order terms to the original quadratic polynomial regression equation, i.e., terms with a total degree of independent variables higher than 2, such as cubic terms, including single-factor cubic terms, interactive cubic terms, etc., to adjust the functional form of the regression equation and improve the model's ability to fit nonlinear response patterns) or add new modeling control factors to adjust the functional form of the quadratic polynomial regression relationship and generate a new quadratic polynomial regression relationship as a new response surface model. Step 4: Perform the lack of fit test and regression significance verification on the new response surface model again. If both the lack of fit test and regression significance verification pass, update the target response surface model based on the new response surface model. Step 5: Perform a misfit check on the updated target response surface model again. If the misfit check passes, save the updated target response surface model. Step 6: If the lack of fit test fails, repeat the lack of fit test failure processing procedure until the preset number of executions is reached and then stop. Preferably, the preset number of executions is 3 times.

[0096] In one embodiment, a water and fertilizer integration optimization control method based on response surface analysis further includes a process for handling regression significance verification failures; The process for handling cases where regression significance verification fails is as follows: Step 1: When the regression significance verification fails, the coding value boundary of the modeling control factor is expanded and adjusted, and the number of trial repetitions is increased. For example, the number of repetitions of the center point is increased to 6-8 times to generate the corrected dimensionless coding baseline parameters. The expansion adjustment rule is to expand and adjust in integer multiples of the unit coding step size, and the expansion range is 1-2 units of coding step size. Step 2: Based on the corrected dimensionless coding baseline parameters, regenerate the coding combination vector containing factorial points, asterisk points, and center points, and convert it into the corresponding water and fertilizer treatment parameters; Step 3: Conduct cultivation experiments based on the regenerated water and fertilizer treatment parameters, collect the measured data of the number of stolons per mother plant or the measured data of the number of daughter plantlets produced by a mother plant, and generate a new experimental sample set; Step 4: Use the least squares method to refit the new experimental sample set with a quadratic polynomial regression relationship to construct a new response surface model; Step 5: Perform the lack of fit test and regression significance verification on the new response surface model again. If both the lack of fit test and regression significance verification pass, update the target response surface model based on the new response surface model. Step 6: Perform regression significance verification on the updated target response surface model again. If the regression significance verification passes, save the updated target response surface model. Step 7: If the regression significance verification fails, repeat the regression significance verification failure handling process until the preset number of executions is reached and then stop.

[0097] Step S140: Decode the optimal combination features of the coded values ​​to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity, and use the optimal single irrigation volume and the optimal irrigation liquid conductivity as the optimal control parameters for fertigation.

[0098] Specifically, after obtaining the optimal combination of coded values, the optimal combination of coded values ​​is decoded to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity. The optimal single irrigation volume and the optimal irrigation liquid conductivity are then used as the optimal control parameters for fertigation.

[0099] Furthermore, to accommodate the control accuracy tolerance of irrigation control equipment in actual production and the flatness of the response surface model near the target solution, this embodiment also sets an operable numerical range, namely the maintenance window. This maintenance window is centered on the stationary point (i.e., the point corresponding to the optimal combination of encoded values). It analyzes the function value changes of the second target response surface model in the neighborhood of the target solution. When the single irrigation volume and irrigation fluid conductivity deviate from the target solution by a certain range (i.e., the maximum range of encoded values ​​within the constraint region, extending outward along each dimension with a fixed step size centered on the stationary point) and the decrease in response value does not exceed a preset decrease tolerance threshold, this range is the maintenance window.

[0100] For example, setting a tolerance threshold for response value decrease. (like =5%), with station ( , Centered on a target, expand outwards along each dimension of the coding space with a fixed step size (e.g., 0.1 coding units) to search for the maximum range of coding values ​​that satisfies the following conditions: and and The physical quantity range corresponding to this maximum encoding value range is the maintenance window.

[0101] In this embodiment, the management window can be dynamically adjusted based on real-time sensor data. For example, soil moisture sensors and conductivity sensors are installed in the cultivation facility to monitor the substrate's moisture content and nutrient concentration in real time. The integrated water and fertilizer control device can receive sensor data via a communication interface (such as a wired serial port or a wireless LoRa module) and compare it with preset irrigation volume management windows and conductivity value management windows. If the real-time monitored value deviates from the management window, the control device automatically adjusts the opening degree of the solenoid valve in the drip irrigation system and the fertilizer absorption ratio of the fertilizer pump to bring the parameters back to near the target solution. For example, when the substrate moisture content is lower than the lower limit of the irrigation volume management window (e.g., the moisture content corresponding to 350 mL / time), the control device can increase the opening time of the solenoid valve to supplement the irrigation volume; when the conductivity value is higher than the upper limit of the conductivity value management window (e.g., 1630 μs / cm), the control device reduces the fertilizer absorption ratio of the fertilizer pump to reduce the nutrient concentration. This dynamic adjustment mechanism can compensate for the impact of environmental changes (such as temperature fluctuations or changes in evaporation rate) on water and fertilizer conditions, thereby maintaining the number of seedlings close to the optimal value.

[0102] Furthermore, in this embodiment, the optimal single irrigation volume, optimal irrigation liquid conductivity, and maintenance window are written into the irrigation program of the integrated water and fertilizer control equipment through the communication interface, thereby controlling the opening degree of the solenoid valve of the drip irrigation system and the fertilizer absorption ratio of the fertilizer pump, so as to achieve precise water and fertilizer supply to the seedlings according to the set parameters.

[0103] The following describes the above-mentioned water and fertilizer integration optimization control method based on response surface analysis with a specific embodiment. However, it should be noted that this specific embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.

[0104] Using the strawberry variety "Hongyan" as the test material, the range of single irrigation volume was set from 229 mL / time to 771 mL / time, and the range of irrigation solution conductivity was set from 500 μs / cm to 1900 μs / cm. Thirteen experimental schemes were generated through a two-factor quadratic general rotatable combination design, including four factorial points, four asterisk points, and five centroid points. This effectively estimated the linear, interaction, and quadratic terms of the model, as well as the experimental error. After cultivation experiments and least squares fitting, specific response surface models (including the first and second response surface models) were obtained. Analysis of variance was performed on the response surface models to verify their reliability and goodness of fit, ensuring that the regression relationships were significant and the lack of fit was not significant. This facilitates the optimization of water and fertilizer integration control parameters to meet the regulatory needs of standardized seedling cultivation.

[0105] This application provides a water and fertilizer integration optimization control method based on response surface methodology. It selects the single irrigation volume and irrigation liquid conductivity as control factors and uses a two-factor quadratic general rotatable combination design to generate multiple factor level combinations. Then, using the encoded values ​​of the single irrigation volume and irrigation liquid conductivity as independent variables, and the number of stolons per mother plant and the number of daughter plants per mother plant as response variables, two quadratic polynomial regression equations are independently fitted using the least squares method based on the same cultivation experiment data, thus obtaining a first and a second original response surface model. To further improve the accuracy of the two original response surface models, after obtaining them, a lack of fit test and regression significance verification are performed. When both the lack of fit test and regression significance verification pass, the original response surface model is determined as the target response surface model. These two target response surface models share the same set of independent variables and experimental data, but their respective regression coefficients are determined by independent least squares fitting, which can capture the different effects of water-fertilizer coupling on the intermediate propagation (number of stolons per mother plant) and the final number of offspring.

[0106] Based on the foregoing embodiments, such as Figure 5 As shown in the figure, this application embodiment further provides a water and fertilizer integration optimization control system based on response surface analysis. This system includes a combined feature generation module, a constraint region construction module, an optimal solution solving module, and a feature decoding module, wherein: The combined feature generation module is used to obtain the single irrigation volume and the conductivity of the irrigation liquid to be processed to form the data to be processed, and to perform dimensionless encoding conversion on the data to be processed to generate the combined features to be processed. The constraint region construction module is used to input the combined features to be processed into the first target response surface model for processing in sequence. It calculates the predicted value of the number of stolons corresponding to each combined feature to be processed in the control factor value space, and constructs the constraint region by combining the predicted value of the number of stolons, the target reproductive channel threshold and the combined features to be processed. The optimal solution solving module is used to sequentially input the combined feature vectors to be processed within the constraint region into the second target response surface model for processing, to obtain the predicted value of the number of seedlings, and to select the combined feature vector to be processed corresponding to the maximum predicted value of the number of seedlings as the optimal encoding value combined feature. The feature decoding module is used to decode the optimal combination of encoded values ​​to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity, which are then used as the optimal control parameters for fertigation.

[0107] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of the invention, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the invention. The sequence numbers of the above-described embodiments of the invention are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0108] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0109] In the several embodiments provided by this invention, it should be understood that the disclosed methods can be implemented in other ways. The methods disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments. The features disclosed in the several method embodiments provided by this invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0110] The above description is merely an embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A water and fertilizer integration optimization control method based on response surface methodology, characterized in that, include: The single irrigation volume and the conductivity of the irrigation solution to be processed are obtained to form data to be processed, and the data to be processed is subjected to dimensionless encoding conversion to generate combined features to be processed. Based on the characteristics of the combination to be processed, the predicted value of the number of stolons in the value space of the control factors is calculated using the first target response surface model. The constrained region is constructed by combining the predicted value of the number of stolons, the target reproductive channel threshold, and the characteristics of the combination to be processed. Within the constrained region, the extreme value solution of the second target response surface model is performed to obtain the optimal combination of encoded values ​​that maximizes the predicted number of seedlings; The optimal combination of encoded values ​​is decoded to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity. The optimal single irrigation volume and the optimal irrigation liquid conductivity are used as the optimal control parameters for fertigation.

2. The water and fertilizer integration optimization control method based on response surface analysis according to claim 1, characterized in that, The construction process of the first target response surface model is as follows: The modeling single irrigation volume and the modeling irrigation liquid conductivity were selected as the modeling control factors, and the water and fertilizer treatment parameters were determined by the point layout rule of the two-factor quadratic general rotational combination design. Obtain the measured data of the number of stolons per mother plant obtained from the cultivation experiment based on the water and fertilizer treatment parameters, and generate the first experimental sample set in combination with the modeling control factors; The first experimental sample set was fitted with a quadratic polynomial regression relationship using the least squares method to construct the first original response surface model. The first original response surface model is subjected to a lack of fit test and a regression significance test. When both the lack of fit test and the regression significance test are passed, the first original response surface model is taken as the first target response surface model.

3. The water and fertilizer integration optimization control method based on response surface methodology according to claim 2, characterized in that, The first target response surface model is: In the formula, This represents the predicted number of runners per individual mother plant. This represents the constant intercept term of the first objective response surface model. This represents the linear effect coefficient of a single irrigation amount on the number of runners per mother plant. This represents the dimensionless coded value corresponding to the single irrigation volume to be processed. This represents the linear effect coefficient of the change in the conductivity of the irrigation solution on the number of runners per mother plant. This represents the dimensionless coded value corresponding to the conductivity of the irrigation solution to be treated. This represents the surface effect coefficient that affects the number of runners per mother plant when the amount of irrigation changes in a single instance. The surface effect coefficient represents the effect of changes in the conductivity of the irrigation solution on the number of runners per mother plant. This represents the interaction term regression coefficient that affects the number of stolons per mother plant when the amount of water used in a single irrigation and the conductivity of the irrigation solution change in tandem.

4. The water and fertilizer integration optimization control method based on response surface analysis according to claim 1, characterized in that, Combining the predicted number of stolons, the target reproductive channel threshold, and the features of the combination to be processed, a constraint region is constructed, including: The target propagation channel threshold is determined based on the planting density of mother seedlings, the target number of daughter seedlings per mu, and the number of daughter seedlings per single stolon. Obtain the predicted number of stolons corresponding to all the combined features to be processed, as well as the coded value boundaries corresponding to the single irrigation volume and the conductivity of the irrigation solution to be processed, and generate a constraint screening judgment benchmark by combining the target reproductive channel threshold. The feasible region of coding values ​​that satisfy the condition that the predicted number of stolons is not lower than the target reproductive channel threshold and that the coding value is within the coding value boundary is determined as the constraint region.

5. The water and fertilizer integration optimization control method based on response surface analysis according to claim 1, characterized in that, The construction process of the second target response surface model is as follows: The modeling single irrigation volume and the modeling irrigation liquid conductivity were selected as the modeling control factors, and the water and fertilizer treatment parameters were determined by the point layout rule of the two-factor quadratic general rotational combination design. Obtain the measured data of the number of daughter seedlings produced by a single mother seedling during the cultivation experiment based on the water and fertilizer treatment parameters, and generate a second experimental sample set by combining the modeling control factors; The least squares method was used to fit the second original response surface model to the first experimental sample set using a quadratic polynomial regression relationship. The second original response surface model is subjected to a lack of fit test and a regression significance test. If both the lack of fit test and the regression significance test pass, the second original response surface model is used as the second target response surface model.

6. The water and fertilizer integration optimization control method based on response surface analysis according to claim 5, characterized in that, The second target response surface model is: In the formula, This represents the predicted number of daughter plantlets produced by a single mother plantlet. This represents the constant intercept term of the second objective response surface model. This represents the linear effect coefficient of a single change in irrigation volume on the number of daughter seedlings produced by a single mother seedling. This represents the dimensionless coded value corresponding to the single irrigation volume to be processed. This represents the linear effect coefficient of the change in the conductivity of the irrigation solution on the number of daughter seedlings produced by a single mother seedling. This represents the dimensionless coded value corresponding to the conductivity of the irrigation solution to be treated. This represents the surface effect coefficient on the number of daughter seedlings produced by a single mother seedling when the amount of irrigation changes. This represents the surface effect coefficient on the number of daughter seedlings produced by a single mother seedling when the conductivity of the irrigation solution changes. This represents the regression coefficient of the interaction term generated by the single mother seedling when the amount of water used for irrigation changes in tandem with the conductivity of the irrigation solution.

7. A water and fertilizer integration optimization control method based on response surface methodology according to claim 2 or 5, characterized in that, The point selection rule for determining water and fertilizer treatment parameters using a two-factor, quadratic, universal rotational combination design includes: Determine the central physical quantity, unit coding step size, and coding value boundary of the modeling control factors, and generate dimensionless coding reference parameters; The point layout rule of the two-factor quadratic general rotation combination design is used to process the dimensionless coding reference parameters to obtain the coding combination vector composed of factor points, asterisk points and center points; Based on the dimensionless coding reference parameters, the coding combination vector is converted into physical quantities to obtain water and fertilizer treatment parameters.

8. A water and fertilizer integration optimization control method based on response surface methodology according to claim 2 or 5, characterized in that, The water and fertilizer integration optimization control method based on response surface analysis also includes a process for handling failures in the lack of fit test. The process for handling failure of the lack of integrability test is as follows: If the lack of conformity test fails, it is confirmed that there is outlier data in the test sample set; If there is abnormal data in the test sample set, the abnormal data is removed, and the test sample set after the abnormal operation is performed is obtained as the corrected sample set. Based on the corrected sample set, a new response surface model is constructed by refitting the quadratic polynomial regression relationship. If there are no outliers in the experimental sample set, add higher-order terms or new modeling control factors to adjust the functional form of the quadratic polynomial regression relationship and generate a new quadratic polynomial regression relationship as a new response surface model. The new response surface model is subjected to a failure test and regression significance verification again. When both the failure test and regression significance verification pass, the target response surface model is updated based on the new response surface model. The updated target response surface model is subjected to a failure test again. If the failure test passes, the updated target response surface model is saved. If the lack of conformity test fails, the lack of conformity test failure processing procedure is repeated until the preset number of executions is reached and then stops.

9. A water and fertilizer integration optimization control method based on response surface analysis according to claim 2 or 5, characterized in that, The water and fertilizer integration optimization control method based on response surface analysis also includes a process for handling cases where regression significance verification fails. The process for handling cases where regression significance verification fails is as follows: When the regression significance verification fails, the coding value boundary of the modeling control factor is expanded and adjusted and the number of trial repetitions is increased to generate the corrected dimensionless coding baseline parameters. Based on the corrected dimensionless coding baseline parameters, a new coding combination vector containing factorial points, asterisk points and center points is generated and converted into the corresponding water and fertilizer treatment parameters. Cultivation experiments were conducted based on the regenerated water and fertilizer treatment parameters. Measured data on the number of stolons per mother plant or the number of daughter plantlets produced by a single mother plant were collected to generate a new experimental sample set. A new response surface model is constructed by refitting the new experimental sample set with a quadratic polynomial regression relationship using the least squares method. The new response surface model is subjected to the lack of fit test and regression significance verification again. When both the lack of fit test and regression significance verification pass, the target response surface model is updated based on the new response surface model. The updated target response surface model is subjected to regression significance verification again. If the regression significance verification passes, the updated target response surface model is saved. If the regression significance verification fails, the regression significance verification failure processing procedure is repeated until the preset number of executions is reached and then stops.

10. A water and fertilizer integration optimization control system based on response surface methodology, characterized in that, The system includes a combined feature generation module, a constraint region construction module, an optimal solution solving module, and a feature decoding module. in: The combined feature generation module is used to obtain the single irrigation volume to be processed and the conductivity of the irrigation liquid to be processed to form the data to be processed, and to perform dimensionless encoding conversion processing on the data to be processed to generate the combined features to be processed. The constraint region construction module is used to calculate the predicted number of stolons in the control factor value space based on the features of the combination to be processed using the first target response surface model, and to construct the constraint region by combining the predicted number of stolons, the target reproductive channel threshold and the features of the combination to be processed. The optimal solution solving module is used to solve the extreme value of the second target response surface model within the constrained area to obtain the optimal combination of encoded values ​​that maximizes the predicted value of the number of seedlings. The feature decoding module is used to decode the optimal combination of encoded values ​​to obtain the optimal single irrigation volume and the optimal irrigation liquid conductivity, and to use the optimal single irrigation volume and the optimal irrigation liquid conductivity as the optimal control parameters for fertigation.

Images