Fish protein amino acid leaf fertilizer preparation regulation and control system based on self-learning optimization

Abstract

The invention provides a fish protein amino acid leaf fertilizer preparation regulation and control system based on self-learning optimization. A hydrolysis characteristic parameter set representing the hydrolysis degree in the enzymolysis reaction process is collected; performing compatibility analysis on hydrolysate generated in the enzymolysis reaction process and a composite auxiliary material in a leaf fertilizer reaction kettle to generate a compatibility quality index; determining a reaction characteristic data set of the sample crop in a state of spraying the leaf surface base solution in a preset period range; inputting the hydrolysis characteristic parameter set, the compatibility quality index and the reaction characteristic data set into a pre-trained self-learning optimization network to extract a recessive association relationship between process parameters of the fish protein amino acid foliar fertilizer and crop reactions; and identifying a process optimization strategy in the preparation process of the foliar fertilizer based on the recessive association relationship, and adjusting the feeding time sequence of the composite auxiliary materials in the enzymolysis reactor according to the process optimization strategy. According to the technical scheme provided by the invention, optimal regulation and control of process parameters can be realized under a whole-process data chain of preparation of the fish protein amino acid foliar fertilizer.

Description

Technical Field

[0001] This application relates to the field of foliar fertilizer preparation regulation technology, and more specifically, this application relates to a fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization. Background Technology

[0002] With the increasing demands of modern agriculture for crop yield and quality, traditional soil fertilization is easily affected by soil pH, compaction, and nutrient fixation, resulting in fertilizer utilization rates of less than 30% and easily causing environmental problems such as eutrophication of water bodies. Foliar fertilizers have become an important supplement due to their advantages such as direct absorption through leaves and rapid onset of action. However, there are technical bottlenecks in the preparation process, such as easy degradation of active ingredients, poor nutrient synergy, and imbalance between leaf adhesion and permeability. Currently, optimizing product performance by controlling the preparation process has become the core direction.

[0003] In the current regulation of foliar fertilizer preparation, the core focus is on "improving nutrient stability and optimizing foliar absorption efficiency," revolving around three dimensions: composition, process, and formulation. Composition regulation involves adjusting pH to inhibit the degradation of active substances such as amino acids and humic acid, ensuring nutrient availability. Process regulation focuses on temperature and stirring rate to balance nutrient solubility and dispersibility with the activity of heat-sensitive components, reducing component failure caused by excessively high local concentrations. Formulation regulation utilizes emulsification or nano-coating technologies to optimize droplet size and enhance leaf adhesion. However, in the regulation of fish protein amino acid foliar fertilizer preparation, existing technologies typically only monitor single parameters such as the degree of hydrolysis during the enzymatic hydrolysis stage or the proportion of excipients added. They fail to establish a complete data chain from raw material enzymatic hydrolysis and formulation synthesis to crop physiological response, resulting in a lack of biological end-point verification for process optimization. Therefore, optimizing and controlling process parameters within the complete data chain of fish protein amino acid foliar fertilizer preparation has become a challenge for the industry. Summary of the Invention

[0004] This application provides a self-learning optimization-based fish protein amino acid foliar fertilizer preparation control system, which can optimize and control process parameters under the whole process data chain of fish protein amino acid foliar fertilizer preparation.

[0005] This application provides a self-learning optimized fish protein amino acid foliar fertilizer preparation control system, the system comprising:

[0006] The acquisition module is used to controllably hydrolyze fish protein raw materials through an enzymatic hydrolysis reactor and to acquire a set of hydrolysis characteristic parameters that characterize the degree of hydrolysis during the enzymatic hydrolysis reaction.

[0007] The compatibility analysis module is used to analyze the compatibility of the hydrolysate generated during the enzymatic hydrolysis process with the composite excipient composed of seaweed polysaccharide, zinc-manganese chelate and humic acid in the foliar fertilizer reaction vessel, and generate the compatibility quality index of the fish protein amino acid foliar fertilizer preparation process.

[0008] The feature data construction module is used to spray the foliar fertilizer base liquid generated in the foliar fertilizer reactor onto the sample crop leaves under controlled light conditions, and to determine the reaction feature dataset of the sample crop under the state of spraying foliar fertilizer base liquid within a preset period range.

[0009] The association extraction module is used to input the hydrolysis feature parameter set, the compatibility quality index and the reaction feature dataset into a pre-trained self-learning optimization network, and the self-learning optimization network extracts the implicit association between the process parameters of fish protein amino acid foliar fertilizer and crop response.

[0010] The adjustment module is used to identify process optimization strategies in the preparation of fish protein amino acid foliar fertilizer based on the implicit correlation, and then adjust the feeding sequence of compound auxiliary materials in the enzymatic hydrolysis reactor according to the process optimization strategies.

[0011] In some embodiments, the hydrolysate generated during the enzymatic hydrolysis process is combined with a composite excipient consisting of seaweed polysaccharide, zinc-manganese chelate, and humic acid in a foliar fertilizer reactor for compatibility analysis. The compatibility quality index of the fish protein amino acid foliar fertilizer preparation process specifically includes:

[0012] The hydrolysate generated during the enzymatic hydrolysis process is added into a foliar fertilizer reaction vessel equipped with a temperature control unit and a stirring unit according to a preset volume ratio.

[0013] The temperature inside the foliar fertilizer reactor is adjusted to a preset range by the temperature control unit and the stirring unit is started. Then, the composite auxiliary materials consisting of seaweed polysaccharide, zinc manganese chelate and humic acid are added according to the preset ratio and stirred continuously until a homogeneous mixed system is formed.

[0014] Extract the set of compatibility physicochemical parameters from the homogeneous mixture system;

[0015] The Analytic Hierarchy Process (AHP) is used to assign weights to each compatibility physicochemical parameter in the set of compatibility physicochemical parameters and to calculate the deviation characteristics from the preset standard parameter thresholds.

[0016] The compatibility quality index of the fish protein amino acid foliar fertilizer preparation process was determined based on all deviation characteristics.

[0017] In some embodiments, the use of the analytic hierarchy process (AHP) to assign weights to each compatibility physicochemical parameter in the compatibility physicochemical parameter set and to calculate the deviation characteristics from preset standard parameter thresholds specifically includes:

[0018] Each compatibility physicochemical parameter in the set of compatibility physicochemical parameters is used as a criterion layer element, and a hierarchical structure model is constructed with compatibility quality evaluation as the target layer.

[0019] Based on the hierarchical structure model, the judgment matrix between each compatibility physicochemical parameter is extracted;

[0020] The consistency test is performed on the judgment matrix. After passing the test, the weights of the corresponding compatibility physicochemical parameters are calculated using the eigenvalue method.

[0021] Calculate the absolute difference between the measured value of each compatibility physicochemical parameter in the compatibility physicochemical parameter set and the corresponding preset standard parameter threshold, and use the absolute difference as the deviation characteristic of the compatibility physicochemical parameter.

[0022] In some embodiments, spraying the foliar fertilizer base solution generated in the foliar fertilizer reactor onto the sample crop leaves under controlled light conditions specifically includes:

[0023] The uniformly growing sample crops were transplanted into the cultivation device in the artificial climate chamber, and the light intensity, photoperiod and temperature and humidity parameters of the artificial climate chamber were set to create a controllable light environment.

[0024] A predetermined volume of foliar fertilizer base solution was extracted from the foliar fertilizer reaction vessel, and the foliar fertilizer base solution was evenly sprayed onto the leaves of the sample crop using a portable sprayer. After spraying, the sample crop was placed in the controlled light environment for reaction monitoring.

[0025] In some embodiments, determining the response characteristic dataset of the sample crop under foliar spraying conditions within a preset period specifically includes:

[0026] The observation start and end times after the sample crops are sprayed with foliar fertilizer base solution are set to define a preset period range, and multiple equally spaced sampling time points are divided within the preset period range;

[0027] At each sampling time point, a multispectral imager was used to scan the leaf surface of the sample crop to obtain multispectral image data containing chlorophyll fluorescence information and leaf morphology details.

[0028] Feature extraction is performed on the multispectral image data to obtain the chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points;

[0029] The chlorophyll fluorescence parameters and leaf morphology parameters at each sampling time point are sequentially correlated and integrated to form a dataset of the response characteristics of the sample crop under the condition of foliar fertilizer base solution application.

[0030] In some embodiments, feature extraction of the multispectral image data to obtain chlorophyll fluorescence parameters and leaf morphology parameters at corresponding sampling time points specifically includes:

[0031] The spectral image at each sampling time point in the multispectral image data is preprocessed to obtain the preprocessed image corresponding to each sampling time point;

[0032] A sampling time point is selected as the selected sampling time point. The threshold segmentation method is used to separate the leaf region from the preprocessed image at the selected sampling time point to obtain a binary image of the leaf region.

[0033] Image data of a preset fluorescence band are extracted from the preprocessed image, and the average fluorescence intensity, fluorescence attenuation rate and photochemical quenching coefficient under the preset fluorescence band are calculated by combining the binarized image of the leaf region to obtain the chlorophyll fluorescence parameters at the selected sampling time point.

[0034] The area, perimeter, and shape factor of the leaf are extracted from the binarized image of the leaf region through morphological operations to obtain the leaf surface morphology parameters at the selected sampling time point.

[0035] Continue to determine the chlorophyll fluorescence parameters and leaf morphology parameters at the remaining sampling time points.

[0036] In some embodiments, identifying process optimization strategies for fish protein amino acid foliar fertilizer preparation based on the implicit association specifically includes:

[0037] The implicit relationships are classified according to the process steps to obtain classified relationship items;

[0038] The classification association items are matched with the preset fertilizer efficiency optimization objective function to screen out key association items whose association strength exceeds the association strength threshold.

[0039] Based on the key correlation items, the adjustment range of the process parameters is determined, and the parameter combinations within the adjustment range are simulated and verified to obtain the fertilizer effect prediction value corresponding to each parameter combination.

[0040] The optimal parameter combination for fertilizer effect prediction was selected as the process optimization strategy in the preparation of fish protein amino acid foliar fertilizer.

[0041] In some embodiments, a set of hydrolysis characteristic parameters is acquired in real time through a sensor array integrated into the enzymatic hydrolysis reactor.

[0042] In some embodiments, the sensor array includes a laser particle size sensor, a high-performance liquid chromatography sensor, an ultraviolet spectrophotometer, and a viscosity sensor.

[0043] In some embodiments, the enzymatic hydrolysis reactor is a closed-loop reaction device that integrates temperature control, dynamic pH balance, stirring and mixing, and online monitoring functions.

[0044] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0045] The self-learning optimization-based fish protein amino acid foliar fertilizer preparation control system provided in this application firstly involves the controllable enzymatic hydrolysis of fish protein raw materials in an enzymatic reactor, and the collection of hydrolysis characteristic parameters characterizing the degree of hydrolysis during the enzymatic hydrolysis reaction. Secondly, the hydrolysate generated during the enzymatic hydrolysis reaction is combined with a composite excipient composed of seaweed polysaccharide, zinc-manganese chelate, and humic acid in a foliar fertilizer reactor for compatibility analysis, generating a compatibility quality index for the fish protein amino acid foliar fertilizer preparation process. Furthermore, the foliar fertilizer base generated in the foliar fertilizer reactor is sprayed onto the leaves of sample crops under controlled light conditions. The process involves: 1) determining the reaction characteristic dataset of the sample crops under foliar spraying conditions within a preset period; 2) inputting the hydrolysis characteristic parameter set, the compatibility quality index, and the reaction characteristic dataset into a pre-trained self-learning optimization network, which then extracts the implicit correlation between the process parameters of the fish protein amino acid foliar fertilizer and the crop response; 3) identifying process optimization strategies in the preparation of the fish protein amino acid foliar fertilizer based on the implicit correlation, and then adjusting the feeding sequence of the compound auxiliary materials in the enzymatic hydrolysis reactor according to the process optimization strategies.

[0046] Therefore, this application demonstrates that the process parameters can be optimized and controlled within a complete data chain for the preparation of fish protein amino acid foliar fertilizer. Firstly, by collecting hydrolysis characteristic parameter sets through an enzymatic hydrolysis reactor, precise monitoring and quantitative characterization of the fish protein hydrolysis process are achieved, providing fundamental data reflecting the degree of hydrolysis for subsequent process optimization. Secondly, compatibility analysis of the hydrolysate and compound excipients in the foliar fertilizer reactor and the generation of a compatibility quality index can quantitatively assess the stability and compatibility of the mixed system, ensuring the basic quality of the foliar fertilizer base solution. Furthermore, spraying the foliar fertilizer base solution under controlled light conditions and determining the reaction characteristic dataset can eliminate external interference and accurately capture the dynamic physiological response of crops to foliar fertilizer, providing direct evidence for verifying the process effect. Then, the hydrolysis characteristic parameter set, compatibility quality index, and reaction characteristics are analyzed together. The self-learning optimization network, pre-trained from the input dataset, extracts latent associations, overcoming the limitations of traditional empirical analysis. It can uncover deep and complex intrinsic connections between process parameters and crop responses, avoiding the problem of insufficient process parameter optimization and control caused by the failure to establish a complete data chain from raw material enzymatic hydrolysis and compound synthesis to crop physiological responses. Finally, based on the latent associations, process optimization strategies are identified and the timing of compound adjuvant additions is adjusted, specifically improving the accuracy and fertilizer efficiency stability of foliar fertilizer preparation processes. This promotes the transformation of foliar fertilizer production from experience-driven to data-driven, achieving process parameter optimization within the complete data chain of fish protein amino acid foliar fertilizer preparation. In summary, the technical solution provided in this application can achieve optimized control of process parameters within the complete data chain of fish protein amino acid foliar fertilizer preparation. Attached Figure Description

[0047] Figure 1 This is a block diagram of a self-learning optimized fish protein amino acid foliar fertilizer preparation regulation system according to some embodiments of this application;

[0048] Figure 2 This is an exemplary flowchart illustrating the determination of the compatibility quality index according to some embodiments of this application. Detailed Implementation

[0049] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] refer to Figure 1 As shown in the figure, this is a modular structure diagram of a fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization according to this embodiment of the present application. The system includes: a data acquisition module 100, a compatibility analysis module 200, a feature data construction module 300, a correlation extraction module 400, and an adjustment module 500, which are described below:

[0051] The acquisition module 100 is used to controllably hydrolyze fish protein raw materials through an enzymatic hydrolysis reactor and to acquire a set of hydrolysis characteristic parameters that characterize the degree of hydrolysis during the enzymatic hydrolysis reaction.

[0052] It should be noted that the enzymatic hydrolysis reactor in this application is a closed-loop reaction device that integrates temperature control, pH dynamic balance, stirring and mixing and online monitoring functions, and is specifically designed to realize the controllable hydrolysis process of proteins (such as fish protein) under the action of enzyme catalysis.

[0053] In practice, firstly, pretreated (e.g., defatted, pulverized) fish protein raw materials are added to the enzymatic hydrolysis reactor at a preset solid-liquid ratio (e.g., 1:5-1:8). The temperature of the reaction system is stabilized within a specific range (e.g., 45-60℃) using a jacketed temperature control system, and the pH value of the reaction environment is maintained by an online pH adjustment module (the specific setting depends on the type of protease used). Secondly, a complex protease (e.g., a mixture of trypsin and papain) is added according to the enzyme-substrate ratio (i.e., 1%-3%), and the stirring device is started to ensure homogeneous mixing of the reaction system. Then, during the enzymatic hydrolysis process, the hydrolysis data are collected in real time by a sensor array integrated into the enzymatic hydrolysis reactor. The feature parameter set, wherein the sensor array includes a laser particle size sensor, a high-performance liquid chromatography (HPLC) sensor, an ultraviolet spectrophotometer, and a viscosity sensor, specifically includes: the average particle size and distribution of protein particles in the hydrolysate monitored by the laser particle size sensor, the concentration and composition of free amino acids detected by the HPLC sensor, the absorbance value at 280 nm measured by the ultraviolet spectrophotometer (reflecting the degree of peptide bond breakage), and the viscosity change of the system recorded in real time by the viscosity sensor. At the same time, the enzymatic hydrolysis time corresponding to each parameter is recorded synchronously by timestamp, forming a hydrolysis feature parameter set characterizing the degree of hydrolysis during the enzymatic hydrolysis reaction.

[0054] It should be noted that the hydrolysis characteristic parameter set in this application refers to a set of dynamic parameters that can quantitatively characterize the degree of protein hydrolysis and product characteristics, which are collected in real time by online monitoring equipment during the fish protein hydrolysis reaction. Obtaining the hydrolysis characteristic parameter set can effectively provide a data basis for adjusting the process parameters of subsequent fish protein amino acid foliar fertilizer preparation.

[0055] The compatibility analysis module 200 is used to analyze the compatibility of the hydrolysate generated during the enzymatic hydrolysis process with the composite excipient composed of seaweed polysaccharide, zinc-manganese chelate and humic acid in the foliar fertilizer reaction vessel, and generate the compatibility quality index of the fish protein amino acid foliar fertilizer preparation process.

[0056] In some embodiments, reference Figure 2As shown in the figure, this is an exemplary flowchart for determining the compatibility quality index according to some embodiments of this application. In this embodiment, the hydrolysate generated during the enzymatic hydrolysis reaction is subjected to compatibility analysis with a composite excipient composed of seaweed polysaccharide, zinc-manganese chelate and humic acid in a foliar fertilizer reactor. The compatibility quality index of the fish protein amino acid foliar fertilizer preparation process can be achieved by the following steps:

[0057] First, in S21, the hydrolysate generated during the enzymatic hydrolysis process is added into the foliar fertilizer reaction vessel equipped with a temperature control unit and a stirring unit according to a preset volume ratio.

[0058] Secondly, in S22, the temperature inside the foliar fertilizer reactor is adjusted to a preset range by the temperature control unit and the stirring unit is started. Then, the composite auxiliary materials consisting of seaweed polysaccharide, zinc manganese chelate and humic acid are added according to the preset ratio and stirred continuously until a homogeneous mixed system is formed.

[0059] Furthermore, in S23, a set of compatibility physicochemical parameters is extracted from the homogeneous mixed system;

[0060] Then, in S24, the Analytic Hierarchy Process (AHP) is used to assign weights to each compatibility physicochemical parameter in the compatibility physicochemical parameter set and to calculate the deviation characteristics from the preset standard parameter threshold.

[0061] Finally, in S25, the compatibility quality index of the fish protein amino acid foliar fertilizer preparation process is determined based on all deviation characteristics.

[0062] In practice, firstly, the hydrolysate generated during the enzymatic hydrolysis reaction is added to a foliar fertilizer reactor equipped with a temperature control unit and a stirring unit according to a preset volume ratio. The temperature control unit can be a jacketed heating device, and the stirring unit can be a paddle stirrer. The foliar fertilizer reactor is a specialized container used in the foliar fertilizer production process to realize core processes such as raw material mixing, dissolution, chemical reaction (such as chelation and compounding), and system homogenization. The preset volume ratio can be set according to actual needs, which will not be elaborated here. Secondly, the temperature inside the foliar fertilizer reactor is adjusted to a preset range suitable for the dissolution and reaction of each auxiliary material by the temperature control unit, and the stirring unit is started at the same time. Then, a composite auxiliary material consisting of seaweed polysaccharide, zinc-manganese chelate, and humic acid is added according to a preset ratio, and stirring is continued until a homogeneous mixed system is formed. The homogeneous mixed system refers to a uniformly mixed liquid of each composite auxiliary material. The preset range and preset ratio can be set according to actual needs. The setup process is not detailed here. Further, online monitoring devices (such as pH sensors, turbidity meters, and chelation rate detectors) are used to collect real-time data on pH, turbidity, and zinc-manganese ion chelation rates from the homogeneous mixture, integrating these data to form a set of compatibility physicochemical parameters. This set of parameters refers to a collection of quantitative parameters that reflect the physical stability and chemical compatibility of the mixture. Then, the analytic hierarchy process (AHP) is used to assign weights to each compatibility physicochemical parameter in the set and calculate the deviation characteristics from preset standard parameter thresholds. These deviation characteristics are quantitative indicators of the degree to which the measured value of a single compatibility physicochemical parameter deviates from the standard threshold. The preset standard parameter thresholds can be set according to actual needs or expert knowledge; no specific limitations are imposed here. Finally, the weights of each compatibility physicochemical parameter and the corresponding deviation characteristics are calculated using a weighted summation method, and the weighted summation result is used as the compatibility quality index for the preparation process of fish protein amino acid foliar fertilizer.

[0063] It should be noted that the compatibility quality index in this application refers to a quantitative indicator that comprehensively measures the degree to which the hydrolysate and the compound excipients meet the performance requirements of foliar fertilizer. By determining the compatibility quality index, the physical stability and chemical compatibility of the hydrolysate mixed with seaweed polysaccharide, zinc-manganese chelate and humic acid compound excipients can be accurately evaluated from a quantitative perspective. This can provide data support for subsequent optimization of excipient ratios and improvement of the overall fertilizer efficiency of foliar fertilizers. At the same time, it can provide quantifiable quality control standards for the standardization and large-scale production of foliar fertilizers, and reduce batch-to-batch quality differences caused by human experience judgment.

[0064] In some embodiments, the following steps can be used to assign weights to each compatibility physicochemical parameter in the compatibility physicochemical parameter set and calculate the deviation characteristics from the preset standard parameter threshold using the analytic hierarchy process:

[0065] Each compatibility physicochemical parameter in the set of compatibility physicochemical parameters is used as a criterion layer element, and a hierarchical structure model is constructed with compatibility quality evaluation as the target layer.

[0066] Based on the hierarchical structure model, the judgment matrix between each compatibility physicochemical parameter is extracted;

[0067] The consistency test is performed on the judgment matrix. After passing the test, the weights of the corresponding compatibility physicochemical parameters are calculated using the eigenvalue method.

[0068] Calculate the absolute difference between the measured value of each compatibility physicochemical parameter in the compatibility physicochemical parameter set and the corresponding preset standard parameter threshold, and use the absolute difference as the deviation characteristic of the compatibility physicochemical parameter.

[0069] In specific implementation, firstly, each compatibility physicochemical parameter in the set of compatibility physicochemical parameters is used as a criterion layer element, and compatibility quality evaluation is used as the target layer. A two-level hierarchical structure model containing the target layer and the criterion layer is constructed. The hierarchical structure model refers to decomposing the complex evaluation problem according to the target, criteria, and other levels to form a logically clear analytical framework. Secondly, based on the hierarchical structure model, the relative importance of each compatibility physicochemical parameter in the criterion layer is compared pairwise according to the 1-9 scaling method (i.e., the general scaling rule of the existing analytic hierarchy process, where 1 indicates that two parameters are equally important, and 9 indicates that one parameter is extremely important than the other). The comparison results are then combined in order into a matrix form to obtain the judgment matrix between each compatibility physicochemical parameter. The judgment matrix refers to the matrix that reflects the relative importance between the compatibility physicochemical parameters. The importance level matrix is ​​then used. Next, existing consistency testing methods are employed to perform consistency testing on the judgment matrix. Once the consistency test is passed, the eigenvector corresponding to the largest eigenvalue of the judgment matrix is ​​solved using the eigenvalue method. The eigenvector is then normalized, and the values ​​in the normalized eigenvector are used as the weights of the corresponding compatibility physicochemical parameters. The weights refer to the importance coefficients of each compatibility physicochemical parameter in the overall compatibility quality evaluation. Finally, the measured values ​​of each compatibility physicochemical parameter are retrieved from the compatibility physicochemical parameter set, and the differences are calculated with pre-set preset standard parameter thresholds (which can be set based on the foliar fertilizer industry quality standards, but are not limited here). The absolute value of the difference is then used as the deviation characteristic of the compatibility physicochemical parameter.

[0070] It should be noted that in this application, deviation characteristics refer to the quantitative index of the degree of deviation between the measured value of a single compatibility physicochemical parameter and the standard threshold. By determining the deviation characteristics, it is possible to directly determine whether the parameter exceeds the allowable fluctuation range, thereby avoiding instability of the compatibility system (such as stratification or degradation of effective components) caused by implicit deviation of parameters, and providing quantitative support for the preliminary screening of compatibility quality.

[0071] The feature data construction module 300 is used to spray the foliar fertilizer base liquid generated in the foliar fertilizer reactor onto the sample crop leaves under controlled light conditions, and to determine the reaction feature dataset of the sample crop under the state of spraying foliar fertilizer base liquid within a preset period.

[0072] In some embodiments, the foliar fertilizer base solution generated in the foliar fertilizer reactor is applied to the sample crop leaves under controlled light conditions using the following steps:

[0073] The uniformly growing sample crops were transplanted into the cultivation device in the artificial climate chamber, and the light intensity, photoperiod and temperature and humidity parameters of the artificial climate chamber were set to create a controllable light environment.

[0074] A predetermined volume of foliar fertilizer base solution was extracted from the foliar fertilizer reaction vessel, and the foliar fertilizer base solution was evenly sprayed onto the leaves of the sample crop using a portable sprayer. After spraying, the sample crop was placed in the controlled light environment for reaction monitoring.

[0075] In practice, firstly, uniformly growing sample crops are transplanted into a cultivation device (containing a cultivation substrate adapted to the crop's native environment) within an artificial climate chamber. The light intensity required for the sample crops is set using the chamber's light control module (such as an LED light source), and the day-night light cycle is set using a time controller. Simultaneously, temperature and humidity sensors, in conjunction with a heating / humidification module, adjust the temperature and humidity parameters within the chamber, creating a controllable light environment that eliminates external environmental interference. This controllable light environment refers to an artificial cultivation environment where environmental factors such as light, temperature, and humidity can be precisely controlled and maintained stably. Secondly, through... The sampling interface of the foliar fertilizer reaction vessel uses a sterile sampling tube to extract a preset volume of foliar fertilizer base solution, which is then poured into a portable sprayer. The portable sprayer is used to evenly spray the foliar fertilizer base solution onto the leaves of the sample crop. After spraying, the sample crop is placed in the controlled light environment for reaction monitoring. The preset volume can be set according to actual needs and is not limited here. The portable sprayer is a liquid storage tank with pressure regulation function and the ability to control droplet size. The reaction monitoring refers to the observation process of continuously tracking the changes in the physiological state and the stability of environmental parameters of the sample crop under the action of foliar fertilizer.

[0076] In some embodiments, determining the response characteristic dataset of the sample crop under foliar spraying conditions within a preset period range is achieved through the following steps:

[0077] The observation start and end times after the sample crops are sprayed with foliar fertilizer base solution are set to define a preset period range, and multiple equally spaced sampling time points are divided within the preset period range;

[0078] At each sampling time point, a multispectral imager was used to scan the leaf surface of the sample crop to obtain multispectral image data containing chlorophyll fluorescence information and leaf morphology details.

[0079] Feature extraction is performed on the multispectral image data to obtain the chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points;

[0080] The chlorophyll fluorescence parameters and leaf morphology parameters at each sampling time point are sequentially correlated and integrated to form a dataset of the response characteristics of the sample crop under the condition of foliar fertilizer base solution application.

[0081] In practice, firstly, combining the growth cycle of the sample crop and the effectiveness of foliar fertilizer, the moment after the foliar fertilizer base solution is sprayed is set as the observation start time, and the moment when the physiological state of the sample crop shows the greatest change is set as the observation end time. These two times together define a preset period range. Then, the preset period range is divided into several sampling time points with equal time intervals. Each sampling time point refers to a specific time node within the preset period used to collect crop physiological data. Secondly, at each sampling time point, a multispectral imager is fixed at a preset distance from the sample crop leaf surface to perform imaging scanning, obtaining multispectral image data containing chlorophyll fluorescence information and leaf morphology details. The preset distance can be adjusted according to actual conditions. Based on actual needs, the multispectral image data refers to a collection of multi-band images that simultaneously contain crop leaf fluorescence information and morphological details. The multispectral images include spectral images of different spectral bands. Then, feature extraction is performed on the multispectral image data to obtain chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points. The chlorophyll fluorescence parameters are quantitative indicators reflecting the crop's photosynthetic capacity, and the leaf morphology parameters are quantitative indicators reflecting the crop's leaf growth status. Finally, according to the chronological order of sampling time, the chlorophyll fluorescence parameters and leaf morphology parameters corresponding to each sampling time point are combined one-to-one into a structured dataset, forming a dataset of the response characteristics of the sample crop under the condition of foliar fertilizer application.

[0082] It should be noted that the response feature dataset in this application refers to a structured data set that records the dynamic changes in the physiological state of sample crops within a preset period. At the data acquisition level, existing technologies mostly collect single-dimensional data such as crop chlorophyll content or leaf morphology, which easily misses key time points for the effectiveness of foliar fertilizers (such as the dynamic changes in fluorescence within a few hours after spraying). However, this solution first defines the preset period by combining the crop growth cycle with the mechanism of action of foliar fertilizers, and then divides the sampling time points at equal intervals, achieving a precise match between the sampling time sequence and the rhythm of crop physiological response. At the same time, a multispectral imager is used to simultaneously acquire chlorophyll fluorescence information and leaf morphology details, avoiding the defect that a single parameter cannot fully reflect the physiological state of crops, forming a "precise time sequence + three-dimensional parameter" acquisition mode. This mode can capture latent physiological changes that are easily overlooked by existing technologies (such as changes in the early chlorophyll fluorescence quenching rate), providing a more complete data foundation for subsequent analysis of crop response.

[0083] In some embodiments, feature extraction of the multispectral image data to obtain the chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points is achieved through the following steps:

[0084] The spectral image at each sampling time point in the multispectral image data is preprocessed to obtain the preprocessed image corresponding to each sampling time point;

[0085] A sampling time point is selected as the selected sampling time point. The threshold segmentation method is used to separate the leaf region from the preprocessed image at the selected sampling time point to obtain a binary image of the leaf region.

[0086] Image data of a preset fluorescence band are extracted from the preprocessed image, and the average fluorescence intensity, fluorescence attenuation rate and photochemical quenching coefficient under the preset fluorescence band are calculated by combining the binarized image of the leaf region to obtain the chlorophyll fluorescence parameters at the selected sampling time point.

[0087] The area, perimeter, and shape factor of the leaf are extracted from the binarized image of the leaf region through morphological operations to obtain the leaf surface morphology parameters at the selected sampling time point.

[0088] Continue to determine the chlorophyll fluorescence parameters and leaf morphology parameters at the remaining sampling time points.

[0089] In specific implementation, firstly, the existing Gaussian filtering algorithm is used to preprocess the spectral image corresponding to each sampling time point in the multispectral image data, resulting in a preprocessed image for each sampling time point. This preprocessed image refers to the spectral image with noise removed. Secondly, one sampling time point is selected from all sampling time points. Using Otsu's method in image processing (which determines the optimal segmentation threshold by calculating the maximum inter-class variance of the image's gray-level histogram), the preprocessed image of the selected sampling time point is divided into two parts: the leaf foreground and the background. This generates a binary image of the leaf region that retains only the leaf region outline. This binary image of the leaf region refers to an image that uses only black and white to represent the leaf region and the background region, respectively. Subsequently, from the preprocessed image of the selected sampling time point, image data of a preset fluorescence band (such as 685nm or 735nm) is extracted based on the characteristic emission band of chlorophyll fluorescence. Combined with the binary image of the leaf region, the effective pixel region of the leaf under the preset fluorescence band is selected. The average fluorescence intensity is obtained by calculating the average gray value of the effective pixel region. Based on the same time at different times... The fluorescence attenuation rate is calculated by measuring the change in fluorescence grayscale value in the fluorescence band. The photochemical quenching coefficient is obtained using the photochemical quenching analysis function of the imaging instrument's software (i.e., calculated based on the fluorescence induction kinetic curve). These three parameters together constitute the chlorophyll fluorescence parameters at the selected sampling time point. The chlorophyll fluorescence parameters are quantitative indicators reflecting the activity of the photosystem during crop photosynthesis. Finally, morphological operations are used to process the binarized image of the leaf region. The total number of pixels in the leaf region is counted using the pixel counting method, and the leaf area is calculated by combining it with the imaging resolution. The leaf contour is extracted using an existing edge detection algorithm, and the pixel length of the contour is calculated and converted to obtain the leaf perimeter. The leaf shape factor is calculated using the formula (shape factor = 4π × area / perimeter squared, π is taken as 3.14). These three parameters together constitute the leaf morphology parameters at the selected sampling time point. The leaf morphology parameters are quantitative indicators reflecting the geometric morphological characteristics of the leaf. Finally, following the same steps as above, the remaining sampling time points are used as new selected sampling time points to continue determining the chlorophyll fluorescence parameters and leaf morphology parameters at all sampling time points.

[0090] The correlation extraction module 400 is used to input the hydrolysis feature parameter set, the compatibility quality index and the reaction feature dataset into a pre-trained self-learning optimization network, and the self-learning optimization network extracts the implicit correlation between the process parameters of fish protein amino acid foliar fertilizer and crop response.

[0091] In some embodiments, the set of hydrolysis feature parameters, the compatibility quality index, and the reaction feature dataset are input into a pre-trained self-learning optimization network, and the self-learning optimization network extracts the implicit correlation between the process parameters of fish protein amino acid foliar fertilizer and crop response using the following steps:

[0092] The hydrolysis characteristic parameter set, compatibility quality index, and reaction characteristic dataset were standardized to obtain a standardized parameter set.

[0093] The standardized parameter set is input into a pre-trained self-learning optimization network. Spatial correlation features and temporal features in the standardized parameter set are extracted through the convolutional and recurrent layers of the network, respectively, and then fused to form a spatiotemporal feature vector.

[0094] The spatiotemporal feature vectors are nonlinearly mapped by the fully connected layer of the network, and the network weights are optimized by combining the backpropagation algorithm to obtain the feature mapping matrix of process parameters and crop response.

[0095] Based on the feature mapping matrix, the implicit correlation between the process parameters of fish protein amino acid foliar fertilizer and crop response was extracted.

[0096] In specific implementation, firstly, the Z-score standardization method (i.e., by calculating the mean and standard deviation of each parameter, converting the parameter values ​​into standardized values ​​with a mean of 0 and a standard deviation of 1) is used to standardize the hydrolysis feature parameter set, the compatibility quality index, and the reaction feature dataset. This eliminates differences in the magnitude of different parameters, resulting in a standardized parameter set. The standardized parameter set refers to a set of parameters in a unified format that eliminates differences in magnitude. Secondly, the standardized parameter set is input into a pre-trained self-learning optimization network. The self-learning optimization network uses pre-set convolutional layers (using 3×3 convolutional kernels) to perform sliding window operations on the standardized parameter set, and the convolutions are then processed. The parameter values ​​within the corresponding local region of the kernel are weighted and summed (the weights can be initially set according to actual needs, which is not limited here), and all weighted sums are combined to form a spatial correlation feature reflecting the spatial distribution of parameters in the standardized parameter set. Simultaneously, the recurrent layer of the self-learning optimization network (i.e., using a long short-term memory structure, using a gating mechanism to remember and update the temporal information in the feature dataset) extracts the temporal features of the parameters in the standardized parameter set as a function of time (that is, the variances of each parameter changing over time are combined to form the temporal features of the parameters in the standardized parameter set as a function of time). Finally, the feature concatenation layer of the self-learning optimization network combines the normalized parameters... The spatial correlation features and the temporal features are combined dimensionally to form a spatiotemporal feature vector containing multi-dimensional information. This spatiotemporal feature vector is a high-dimensional vector that simultaneously contains both spatial correlation information and temporal variation information. Then, the spatiotemporal feature vector is transformed into low-dimensional features directly related to the process parameters and crop responses in the preparation of fish protein amino acid foliar fertilizer by using a self-learning optimization network's fully connected layers (i.e., composed of multiple neurons that perform linear transformations and nonlinear mappings on the input vector through activation functions). Simultaneously, the network weights are iteratively optimized using a backpropagation algorithm (i.e., by calculating the error between the network's predicted output and the actual target value, adjusting the weight parameters of each layer in reverse along the network hierarchy). The weight parameters of the fully connected layer and the entire network are optimized, and the mapping relationship between the network and the crop response characteristics in this state is stored in matrix form to obtain the feature mapping matrix between the process parameters and the crop response. The feature mapping matrix is ​​a matrix that reflects the correspondence between the process parameters and the crop response in various dimensions. Finally, the correlation coefficient of each element in the feature mapping matrix is ​​calculated by the Pearson correlation coefficient method, and the associated elements with correlation coefficients exceeding a preset threshold are screened out. These associated elements are then integrated to form the implicit correlation between the process parameters of fish protein amino acid foliar fertilizer and the crop response. The preset threshold can be set according to actual needs and is not limited here.

[0097] It should be noted that during the training of the pre-trained self-learning optimization network, the historical hydrolysis characteristic parameter set, compatibility quality index, and corresponding crop response characteristic dataset of the foliar fertilizer preparation process are first collected. The training set and validation set are divided proportionally and standardized. Then, the network structure (including convolutional layers, recurrent layers, and fully connected layers) is initialized, and the loss function (such as the mean squared error function), optimizer (such as stochastic gradient descent), and learning rate are set. Subsequently, the standardized training set data is input into the network. Spatial features are extracted by the convolutional layer, and temporal features are extracted by the recurrent layer and fused into a spatiotemporal feature vector. The prediction result is then output through nonlinear mapping by the fully connected layer. The loss value is calculated by comparing the prediction result with the actual crop response label. The weights of each layer of the network are updated by taking derivatives layer by layer using the backpropagation algorithm. At the same time, the network performance is evaluated using the validation set after each iteration. The training is iterated repeatedly until the loss value of the validation set converges and stabilizes (without significant fluctuations or decreases). The training is then stopped and the network parameters at this time are saved, thus completing the training of the pre-trained self-learning optimization network. The resulting network can be used for feature extraction and association mining of subsequent new data.

[0098] It should also be noted that the implicit correlation in this application refers to the inherent correspondence between process parameters and crop response that is not directly apparent. In terms of correlation mining, existing technologies for correlation analysis between foliar fertilizer processes and crop responses mostly remain at the level of linear correspondence between "single process parameter (such as enzymatic hydrolysis time) and single crop index (such as chlorophyll content)," failing to cover the complex interactive relationships between hydrolysis feature parameter sets, compatibility quality indices, and response feature datasets. This solution, however, standardizes three types of heterogeneous data and inputs them into a self-learning optimization network. It uses convolutional layers to extract spatial correlations between parameters and recurrent layers to capture temporal dynamic correlations, and then fuses them to form a spatiotemporal feature vector. This achieves cross-dimensional correlation mining of "multi-source parameters - spatiotemporal dimension - crop response," breaking through the analysis bottleneck of single-dimensional, static correlation in existing technologies.

[0099] The adjustment module 500 is used to identify process optimization strategies in the preparation process of fish protein amino acid foliar fertilizer based on the implicit correlation, and then adjust the feeding sequence of compound auxiliary materials in the enzymatic hydrolysis reactor according to the process optimization strategies.

[0100] In some embodiments, the process optimization strategy for identifying fish protein amino acid foliar fertilizer preparation based on the implicit association is achieved through the following steps:

[0101] The implicit relationships are classified according to the process steps to obtain classified relationship items;

[0102] The classification association items are matched with the preset fertilizer efficiency optimization objective function to screen out key association items whose association strength exceeds the association strength threshold.

[0103] Based on the key correlation items, the adjustment range of the process parameters is determined, and the parameter combinations within the adjustment range are simulated and verified to obtain the fertilizer effect prediction value corresponding to each parameter combination.

[0104] The optimal parameter combination for fertilizer effect prediction was selected as the process optimization strategy in the preparation of fish protein amino acid foliar fertilizer.

[0105] In practice, firstly, based on the technological steps of foliar fertilizer preparation (such as enzymatic hydrolysis and compatibility steps) and crop response indicators (such as chlorophyll fluorescence parameters and leaf morphology parameters), implicit correlations are categorized and sorted. Correspondences involving parameters such as enzymatic hydrolysis temperature and time are classified into the enzymatic hydrolysis process category, while correlations involving parameters such as excipient ratio and reactor temperature are classified into the compatibility process category, resulting in categorized correlation items. These categorized correlation items refer to subsets of implicit correlations after being divided according to technological steps. Secondly, a pre-defined fertilizer efficiency optimization objective function is constructed (i.e., with the improvement rate of key crop physiological indicators and the retention rate of effective components in foliar fertilizer as core objectives, transforming multiple objectives into a single objective function through weighted summation). The correlation strength of each correlation in the categorized correlation items (i.e., the previously calculated correlation coefficient) is multiplied by the weight coefficient of the objective function. Correspondences with a weighted correlation strength exceeding a critical correlation strength value are selected as key correlation items. The critical correlation strength value can be based on industry benchmarks for foliar fertilizers. The minimum requirements for improving efficiency are determined and will not be elaborated here. The key correlations refer to implicit correlations that significantly contribute to the fertilizer efficiency optimization target. Then, based on the influence of process parameters and crop response in the key correlations (e.g., increased enzymatic hydrolysis temperature can improve amino acid conversion rate, but excessive temperature will lead to degradation of effective components), combined with the equipment capacity and raw material characteristics of foliar fertilizer production, adjustment ranges for each process parameter are defined (e.g., temperature adjustment range, auxiliary material dosage range). Orthogonal experimental design is used to simulate and verify the parameter combinations within the adjustment range. That is, the parameter combinations are substituted into the fertilizer efficiency prediction model (i.e., a multiple regression model constructed based on historical production data and crop response data) to calculate the fertilizer efficiency prediction value corresponding to each parameter combination. The fertilizer efficiency prediction value refers to the quantitative result reflecting the expected fertilizer efficiency of the parameter combination. Finally, the fertilizer efficiency prediction values ​​of all parameter combinations are compared, and the parameter combination with the highest fertilizer efficiency prediction value is selected as the process optimization strategy in the preparation of fish protein amino acid foliar fertilizer.

[0106] It should be noted that the process optimization strategy in this application refers to the optimization scheme for adjusting specific parameters during the preparation of fish protein amino acid foliar fertilizer. The core function of determining the process optimization strategy during the preparation of fish protein amino acid foliar fertilizer is to transform the inherent laws between process parameters and crop response that were previously discovered through implicit correlations into specific parameter adjustment schemes that can directly guide actual production, thereby solving the problem of process parameters relying on experience setting and the ambiguity of optimization direction in traditional foliar fertilizer production.

[0107] In some embodiments, adjusting the feeding sequence of the composite excipients in the enzymatic hydrolysis reactor according to the process optimization strategy is achieved through the following steps:

[0108] The parameter adjustment instructions related to the composite excipients in the process optimization strategy were analyzed, and the optimal feeding nodes and intervals for seaweed polysaccharide, zinc-manganese chelate and humic acid were determined.

[0109] The optimal feeding node is matched with the set of hydrolysis characteristic parameters monitored in real time in the enzymatic hydrolysis reactor in order to set the trigger threshold corresponding to each auxiliary material.

[0110] When the hydrolysis characteristic parameters reach the preset trigger threshold during the enzymatic hydrolysis reaction, the automatic feeding module of the enzymatic hydrolysis reactor sequentially feeds seaweed polysaccharide, zinc-manganese chelate and humic acid at corresponding intervals.

[0111] In specific implementation, firstly, parameter adjustment instructions related to the composite excipients (seaweed polysaccharide, zinc-manganese chelate, humic acid) are extracted from the process optimization strategy. Through instruction parsing, the optimal time point for each excipient to be added during the enzymatic hydrolysis reaction and the time interval between the addition of two adjacent excipients are determined. The optimal addition point refers to the specific reaction moment that maximizes the reaction effect between the excipient and the hydrolysate, and the interval refers to the time difference between the addition of the two excipients. Secondly, the determined optimal addition point is mapped and matched with the set of hydrolysis characteristic parameters monitored in real time by sensors in the enzymatic hydrolysis reactor. That is, a threshold setting method (based on the parameter value range corresponding to the optimal addition point in historical production data) is used to set a critical value for each excipient that triggers the addition action when the hydrolysis characteristic parameter reaches that value (i.e., the optimal addition point is set as the optimal addition point). The minimum parameter value corresponding to the feeding node is used as the critical value to obtain the trigger threshold for each auxiliary material. The trigger threshold refers to the critical value of the hydrolysis characteristic parameter for starting the feeding of the auxiliary material. Finally, during the enzymatic hydrolysis reaction, the hydrolysis characteristic parameter is continuously collected by the real-time monitoring system of the enzymatic hydrolysis reactor and compared with the preset trigger threshold. When the hydrolysis characteristic parameter corresponding to a certain auxiliary material reaches the corresponding trigger threshold, the automatic feeding module of the enzymatic hydrolysis reactor (composed of a metering pump and feeding pipeline driven by a PLC control system) sequentially and quantitatively feeds seaweed polysaccharide, zinc manganese chelate and humic acid into the enzymatic hydrolysis reactor at set intervals to ensure that the auxiliary material participates in the enzymatic hydrolysis process in the optimal reaction stage. The trigger threshold can be set according to actual needs or according to expert knowledge, and is not limited here.

[0112] It should be noted that, in this embodiment, the automatic feeding module refers to a device system that automatically completes the addition of auxiliary materials according to a preset program and triggering conditions.

[0113] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0114] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A self-learning optimized foliar fertilizer preparation control system based on fish protein amino acids, characterized in that, The system includes: The acquisition module is used to controllably hydrolyze fish protein raw materials through an enzymatic hydrolysis reactor and to acquire a set of hydrolysis characteristic parameters that characterize the degree of hydrolysis during the enzymatic hydrolysis reaction. The compatibility analysis module is used to analyze the compatibility of the hydrolysate generated during the enzymatic hydrolysis process with the composite excipient composed of seaweed polysaccharide, zinc-manganese chelate and humic acid in the foliar fertilizer reaction vessel, and generate the compatibility quality index of the fish protein amino acid foliar fertilizer preparation process. The feature data construction module is used to spray the foliar fertilizer base liquid generated in the foliar fertilizer reactor onto the sample crop leaves under controlled light conditions, and to determine the reaction feature dataset of the sample crop under the state of spraying foliar fertilizer base liquid within a preset period range. The association extraction module is used to input the hydrolysis feature parameter set, the compatibility quality index and the reaction feature dataset into a pre-trained self-learning optimization network, and the self-learning optimization network extracts the implicit association between the process parameters of fish protein amino acid foliar fertilizer and crop response. The adjustment module is used to identify process optimization strategies in the preparation of fish protein amino acid foliar fertilizer based on the implicit correlation, and then adjust the feeding sequence of compound auxiliary materials in the enzymatic hydrolysis reactor according to the process optimization strategies.

2. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, The hydrolysate generated during the enzymatic hydrolysis process was mixed with a composite excipient consisting of seaweed polysaccharides, zinc-manganese chelates, and humic acid in a foliar fertilizer reactor for compatibility analysis. The compatibility quality index of the fish protein amino acid foliar fertilizer preparation process specifically includes: The hydrolysate generated during the enzymatic hydrolysis process is added into a foliar fertilizer reaction vessel equipped with a temperature control unit and a stirring unit according to a preset volume ratio. The temperature inside the foliar fertilizer reactor is adjusted to a preset range by the temperature control unit and the stirring unit is started. Then, the composite auxiliary materials consisting of seaweed polysaccharide, zinc manganese chelate and humic acid are added according to the preset ratio and stirred continuously until a homogeneous mixed system is formed. Extract the set of compatibility physicochemical parameters from the homogeneous mixture system; The Analytic Hierarchy Process (AHP) is used to assign weights to each compatibility physicochemical parameter in the set of compatibility physicochemical parameters and to calculate the deviation characteristics from the preset standard parameter thresholds. The compatibility quality index of the fish protein amino acid foliar fertilizer preparation process was determined based on all deviation characteristics.

3. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 2, characterized in that, The analytic hierarchy process (AHP) is used to assign weights to each compatibility physicochemical parameter in the set and to calculate the deviation characteristics from preset standard parameter thresholds. Specifically, this includes: Each compatibility physicochemical parameter in the set of compatibility physicochemical parameters is used as a criterion layer element, and a hierarchical structure model is constructed with compatibility quality evaluation as the target layer. Based on the hierarchical structure model, the judgment matrix between each compatibility physicochemical parameter is extracted; The consistency test is performed on the judgment matrix. After passing the test, the weights of the corresponding compatibility physicochemical parameters are calculated using the eigenvalue method. Calculate the absolute difference between the measured value of each compatibility physicochemical parameter in the compatibility physicochemical parameter set and the corresponding preset standard parameter threshold, and use the absolute difference as the deviation characteristic of the compatibility physicochemical parameter.

4. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, The foliar fertilizer base solution generated in the foliar fertilizer reaction vessel was sprayed onto the sample crops under controlled light conditions, specifically including: The uniformly growing sample crops were transplanted into the cultivation device in the artificial climate chamber, and the light intensity, photoperiod and temperature and humidity parameters of the artificial climate chamber were set to create a controllable light environment. A predetermined volume of foliar fertilizer base solution was extracted from the foliar fertilizer reaction vessel, and the foliar fertilizer base solution was evenly sprayed onto the leaves of the sample crop using a portable sprayer. After spraying, the sample crop was placed in the controlled light environment for reaction monitoring.

5. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, The dataset of response characteristics of the sample crops under foliar spraying conditions within a preset period specifically includes: The observation start and end times after the sample crops are sprayed with foliar fertilizer base solution are set to define a preset period range, and multiple equally spaced sampling time points are divided within the preset period range; At each sampling time point, a multispectral imager was used to scan the leaf surface of the sample crop to obtain multispectral image data containing chlorophyll fluorescence information and leaf morphology details. Feature extraction is performed on the multispectral image data to obtain the chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points; The chlorophyll fluorescence parameters and leaf morphology parameters at each sampling time point are sequentially correlated and integrated to form a dataset of the response characteristics of the sample crop under the condition of foliar fertilizer base solution application.

6. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 5, characterized in that, Feature extraction of the multispectral image data to obtain chlorophyll fluorescence parameters and leaf morphology parameters at the corresponding sampling time points specifically includes: The spectral image at each sampling time point in the multispectral image data is preprocessed to obtain the preprocessed image corresponding to each sampling time point; A sampling time point is selected as the selected sampling time point. The threshold segmentation method is used to separate the leaf region from the preprocessed image at the selected sampling time point to obtain a binary image of the leaf region. Image data of a preset fluorescence band are extracted from the preprocessed image, and the average fluorescence intensity, fluorescence attenuation rate and photochemical quenching coefficient under the preset fluorescence band are calculated by combining the binarized image of the leaf region to obtain the chlorophyll fluorescence parameters at the selected sampling time point. The area, perimeter, and shape factor of the leaf are extracted from the binarized image of the leaf region through morphological operations to obtain the leaf surface morphology parameters at the selected sampling time point. Continue to determine the chlorophyll fluorescence parameters and leaf morphology parameters at the remaining sampling time points.

7. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, The specific process optimization strategies for fish protein amino acid foliar fertilizer preparation based on the aforementioned implicit correlation include: The implicit relationships are classified according to the process steps to obtain classified relationship items; The classification association items are matched with the preset fertilizer efficiency optimization objective function to screen out key association items whose association strength exceeds the association strength threshold. Based on the key correlation items, the adjustment range of the process parameters is determined, and the parameter combinations within the adjustment range are simulated and verified to obtain the fertilizer effect prediction value corresponding to each parameter combination. The optimal parameter combination for fertilizer effect prediction was selected as the process optimization strategy in the preparation of fish protein amino acid foliar fertilizer.

8. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, A sensor array integrated into the enzymatic hydrolysis reactor is used to collect a set of hydrolysis characteristic parameters in real time.

9. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 8, characterized in that, The sensor array includes a laser particle size sensor, a high-performance liquid chromatography sensor, an ultraviolet spectrophotometer, and a viscosity sensor.

10. The fish protein amino acid foliar fertilizer preparation regulation system based on self-learning optimization as described in claim 1, characterized in that, The enzymatic hydrolysis reactor is a closed-loop reaction device that integrates temperature control, dynamic pH balance, stirring and mixing, and online monitoring functions.

Images