Method and system for optimally proportioning microbial fertilizer protective agent and preservative

Abstract

The invention discloses a method and a system for optimally proportioning a microbial fertilizer protective agent and a preservative. The method is applied to a system comprising a cloud computing server, an edge computing node and a laboratory detection unit, environmental parameters such as temperature, humidity and pH value and bacterial activity data of the microbial fertilizer are collected through the edge computing node, and the data are uploaded to the cloud computing server. And the cloud computing server constructs a formula optimization model based on the collected data, calculates and outputs an optimal proportioning scheme, and controls the edge computing node to monitor and verify the bacterium activity stability of the optimal proportioning scheme. And dynamically adjusting the proportioning parameter of the protective agent and the preservative according to the verification result. The system integrates a high-precision sensor, data acquisition equipment and laboratory detection equipment, so that intelligent optimization of the ratio of the microbial fertilizer protective agent to the preservative is realized. The bacterial activity stability of the microbial fertilizer can be effectively improved, and the storage life is prolonged.

Description

Technical Field

[0001] The present application relates to the technical field of microbial fertilizers, and in particular to a method and system for optimizing the ratio of a microbial fertilizer protective agent and a preservative based on environmental conditions and bacterial activity. Background Art

[0002] With the development of microbial fertilizer technology, how to ensure that microbial fertilizers maintain stable bacterial activity during storage and transportation has become an important issue. The effective bacteria in microbial fertilizers are easily inactivated by environmental factors, which seriously affects the effect and use value of the product.

[0003] At present, the empirical formulation method is mainly used, that is, the type and dosage of protective agents and preservatives are selected based on experience. This method is usually carried out in small batches under laboratory conditions and then directly applied to large-scale production. Another method is to use a single environmental parameter monitoring, focusing only on the effect of temperature or humidity on bacterial activity.

[0004] The most relevant existing technology is to use data collection and analysis methods to analyze the effects of different formulations on microbial activity by collecting environmental parameters and bacterial activity data. However, this method has the following problems: first, data collection is not systematic and complete, and often only focuses on test results under certain specific conditions; second, there is a lack of effective data analysis models, and it is impossible to accurately predict the changes in bacterial activity under different conditions; third, the formulation optimization process lacks a dynamic adjustment mechanism, making it difficult to adapt to the needs of different storage environments. Summary of the invention

[0005] The present application provides a method and system for optimizing the ratio of a microbial fertilizer protective agent and a preservative, which solves the problems in the prior art of unsystematic data collection, inaccurate analysis model, and lack of dynamic adjustment mechanism for formula optimization.

[0006] The present application embodiment provides a method for optimizing the ratio of a microbial fertilizer protective agent and a preservative, comprising:

[0007] Collect microbial data through edge computing nodes, where the microbial data includes temperature, humidity, and pH parameters of the storage environment, as well as bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions;

[0008] The edge computing node uploads the collected microbial data to a cloud computing server;

[0009] The cloud computing server constructs a formula optimization model based on the temperature, humidity, pH value parameters and the bacterial activity data;

[0010] The cloud computing server calculates and outputs the optimal ratio scheme according to the formula optimization model, and controls the edge computing node to monitor and verify the bacterial activity stability of the optimal ratio scheme under different storage conditions;

[0011] The cloud computing server receives the bacterial activity detection data regularly uploaded by the edge computing node, performs stability analysis, and dynamically adjusts the ratio parameters of the protective agent and the preservative according to the analysis results.

[0012] Collecting microbial data through the edge computing node, the microbial data including temperature, humidity and pH parameters of the storage environment, and bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions, wherein the collection range of the temperature parameter is 20-40°C, the collection range of the humidity parameter is 0-100% relative humidity, the collection range of the pH parameter is 3-9, and the bacterial activity data includes monthly colony counts and bacterial survival rates, which are collected by the laboratory detection unit;

[0013] The edge computing node uploads the collected microbial data to a cloud computing server;

[0014] The cloud computing server constructs a formula optimization model based on the temperature, humidity, pH value parameters and the bacterial activity data;

[0015] The cloud computing server calculates and outputs the optimal ratio scheme according to the formula optimization model, and controls the edge computing node to monitor and verify the bacterial activity stability of the optimal ratio scheme under different storage conditions, wherein the optimal ratio scheme has a protective agent concentration range of 0-5‰ and a preservative concentration range of 0-2‰;

[0016] The cloud computing server receives the bacterial activity detection data regularly uploaded by the edge computing node, performs stability analysis, and dynamically adjusts the ratio parameters of the protective agent and the preservative according to the analysis results.

[0017] Optionally, the microbial data is collected through the edge computing node, and the microbial data includes bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions, including:

[0018] The colony counts of the two basic formulations, 20-20-20+TE and 12-6-40+TE, were performed using the standard plate count method;

[0019] In each basic formula, Bacillus subtilis and Bacillus amyloliquefaciens were mixed in a ratio of 1:1, and the initial amount of bacteria added was 4.00E+08 CFU / g;

[0020] Monitor colony counts and pH changes monthly;

[0021] Analyze bacterial survival under different pH conditions;

[0022] Correlation data between environmental condition parameters and bacterial activity were recorded.

[0023] Optionally, the recipe optimization model includes a high-dimensional function approximation model, and constructing the recipe optimization model includes:

[0024] Temperature, humidity, pH value, storage time, protective agent concentration, and preservative concentration were used as input variables to construct a high-dimensional feature space;

[0025] Identify effective dimensions that significantly affect bacterial activity from the input variables;

[0026] Based on the effective dimensions, the model complexity is reduced to construct a simplified model;

[0027] Verifying the prediction accuracy of the simplified model through experimental data;

[0028] The simplified model parameters are optimized according to the verification results.

[0029] Optionally, identifying effective dimensions that significantly affect bacterial activity from the input variables, reducing model complexity based on the effective dimensions, and constructing a simplified model includes:

[0030] Determine temperature as the first dominant dimension and establish the temperature influence function;

[0031] The correlation between pH value and bacterial activity was analyzed, and pH value was determined as the second dominant dimension;

[0032] Establish a progressive decay model of storage time as the third dominant dimension;

[0033] The simplified model is constructed by integrating three dominant dimensions;

[0034] The simplified model parameters are calculated based on experimental data.

[0035] Optionally, calculating and outputting an optimal ratio scheme according to the recipe optimization model includes:

[0036] Classify the bacterial activity data under different formulation conditions and generate bacterial activity data sets within different pH ranges;

[0037] Based on the bacterial activity data set, a two-dimensional distribution function is generated on a pH-bacterial activity plane and data projection is performed to obtain projection data;

[0038] Performing Radon transformation on the projection data to obtain a transformed distribution function;

[0039] A regularization term is introduced into the transformed distribution function to deal with the uncertainty caused by the limited number of samples, and a regularized distribution estimate is obtained;

[0040] Based on the regularized distribution estimation, a probability distribution model for predicting bacterial activity levels under different pH conditions is constructed;

[0041] The optimal matching scheme is calculated and output according to the probability distribution model.

[0042] Optionally, constructing a probability distribution estimation model includes:

[0043] Estimate the probability distribution of maintaining the target bacterial viability level under normal temperature conditions, wherein the normal temperature condition is 25°C;

[0044] Estimating the probability distribution of maintaining the target bacterial viability level under high temperature conditions, wherein the high temperature condition is 40°C;

[0045] Establish classification decision rules for recipe optimization based on probability distribution;

[0046] Verify the accuracy of classification decision rules based on experimental data;

[0047] Dynamically adjust probability distribution model parameters.

[0048] Optionally, the stability analysis includes: verifying the bacterial activity stability of the optimal ratio solution under different storage conditions, specifically including:

[0049] Store at room temperature and humidity for 4 months, and test bacterial activity once a month. The room temperature and humidity conditions are 25°C and 45% relative humidity.

[0050] Store at room temperature and high humidity for 4 months, and test the bacterial activity once a month. The room temperature and high humidity conditions are 25°C and 85% relative humidity;

[0051] Store under high temperature and normal humidity conditions for 4 months, and test the bacterial activity once a month. The high temperature and normal humidity conditions are 40°C and 45% relative humidity;

[0052] Store under high temperature and high humidity conditions for 4 months, and test the bacterial activity once a month. The high temperature and high humidity conditions are 40°C and 85% relative humidity;

[0053] Record and analyze the changing trends of bacterial activity under different storage conditions.

[0054] Optionally, the bacterial activity is tested once a month, including:

[0055] The total number of colonies was determined by the standard plate count method;

[0056] Calculate the retention rate relative to the initial bacterial activity;

[0057] Determine the current pH value and calculate its change;

[0058] Analyze the remaining effective content of protective agents and preservatives.

[0059] Optionally, the dynamically adjusting parameters of the ratio of the protective agent to the preservative includes:

[0060] Analyze the deviation between bacterial activity test results and expected targets;

[0061] Evaluate the effect of pH stability on bacterial activity;

[0062] Adjust the amount of protective agent added according to the evaluation results;

[0063] Adjust the amount of preservatives added based on the evaluation results;

[0064] Update relevant parameters in the recipe optimization model.

[0065] The present application also provides a microbial fertilizer protective agent and preservative optimization ratio system, which is used to implement the above-mentioned microbial fertilizer protective agent and preservative optimization ratio method, and the system includes:

[0066] Edge computing nodes, including:

[0067] Digital temperature sensor with an accuracy of ±0.1°C and an operating range of 20-40°C;

[0068] Capacitive humidity sensor with an accuracy of ±1%RH and a detection range of 0-100%;

[0069] Digital pH sensor with an accuracy of ±0.01pH and a measurement range of 3-9pH;

[0070] Signal conditioning circuits for sensor data acquisition;

[0071] A microprocessor for data preprocessing;

[0072] Industrial-grade programmable controller;

[0073] Industrial Ethernet communication module;

[0074] Edge computing processor using ARM architecture;

[0075] Cloud computing servers, including:

[0076] Computing servers for data analysis;

[0077] Distributed database for data storage;

[0078] Laboratory testing unit, including:

[0079] Automatic colony counting system;

[0080] Digital microscope with image analysis capabilities;

[0081] a spectrophotometer for measuring bacterial density;

[0082] Environmental test chamber with controlled temperature and humidity;

[0083] Sterile operating table.

[0084] The present application also provides a computer device, the computer device comprising:

[0085] at least one processor; and,

[0086] a memory communicatively connected to the at least one processor; wherein,

[0087] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for optimizing the ratio of the microbial fertilizer protective agent and the preservative.

[0088] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the method for optimizing the ratio of the microbial fertilizer protective agent and the preservative.

[0089] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method for optimizing the ratio of the above-mentioned microbial fertilizer protective agent and the preservative.

[0090] This application has the following technical effects:

[0091] By establishing a systematic data collection system, comprehensive monitoring of microbial fertilizer storage environment parameters and bacterial activity data is achieved, providing a reliable data basis for formula optimization.

[0092] An accurate recipe optimization model was constructed by using a high-dimensional function approximation model and Radon cumulative distribution transformation method, which improved the accuracy of the optimization results.

[0093] A dynamic adjustment mechanism based on real-time monitoring data has been established, which enables the ratio of protective agents and preservatives to be optimized according to actual storage conditions, thereby improving the storage stability of microbial fertilizers. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0095] Figure 1 A schematic diagram of the overall architecture of the system for optimizing the ratio of microbial fertilizer protective agent and preservative provided in the embodiment of the present application;

[0096] Figure 2 A schematic diagram of the structure of an edge computing node provided in an embodiment of the present application;

[0097] Figure 3 A schematic diagram of the data collection process provided in the embodiment of the present application;

[0098] Figure 4 A schematic diagram of the formulation optimization model construction process provided in the embodiment of the present application;

[0099] Figure 5 A schematic diagram of the stability verification process provided in the embodiment of the present application;

[0100] Figure 6 A schematic diagram of a process for dynamically adjusting the ratio of protective agent to preservative provided in an embodiment of the present application;

[0101] Figure 7 The first live bacteria count experiment result under normal temperature and humidity conditions provided in the examples of this application;

[0102] Figure 8a The second live bacteria count experiment result under normal temperature and humidity conditions provided in the examples of this application;

[0103] Figure 8b The second live bacteria count experiment result under normal temperature and high humidity conditions provided in the examples of this application;

[0104] FIG8c is a second live bacteria counting experiment result under high temperature and high humidity conditions provided in an embodiment of the present application;

[0105] Fig. 9 The experimental results after optimizing the ratio provided in the embodiments of the present application are shown. DETAILED DESCRIPTION

[0106] The specific implementation methods of the present application are described in detail below with reference to the accompanying drawings.

[0107] See also Figure 1 The microbial fertilizer protective agent and preservative optimization ratio system provided in the embodiment of the present application includes a cloud computing server, an edge computing node and a laboratory detection unit.

[0108] Before describing step S1 in detail, in order to allow those skilled in the art to better understand the present application, the terms involved in the present application are first explained.

[0109] "Bacterial activity data" refers to the survival of effective bacteria in microbial fertilizers, which is mainly expressed by colony counts (CFU / g).

[0110] "Formula optimization model" refers to a mathematical model used to predict and optimize the ratio of protective agents to preservatives, which is built based on environmental parameters and bacterial activity data.

[0111] S1: Overall steps

[0112] The method for optimizing the ratio of microbial fertilizer protective agent and preservative provided in the present application is applied to a system including a cloud computing server, an edge computing node and a laboratory detection unit. Among them, the edge computing node is responsible for collecting microbial data, including the temperature, humidity and pH parameters of the storage environment, and the bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions. Specifically:

[0113] S1.1 Data collection phase:

[0114] Microbial data is collected through edge computing nodes, with the temperature parameter collection range of 20-40°C, the humidity parameter collection range of 0-100% relative humidity, and the pH value parameter collection range of 3-9. Bacterial activity data includes monthly colony counts and bacterial survival rates, which are collected by laboratory testing units.

[0115] Among them, the edge computing node includes multiple sensor modules, including:

[0116] 1) Temperature sensor module: It uses a digital temperature sensor with an accuracy of ±0.1°C, an operating range of 20-40°C, and a sampling frequency of once every 30 minutes. It is used to monitor the temperature changes of the storage environment in real time.

[0117] 2) Humidity sensor module: It uses a capacitive humidity sensor with an accuracy of ±1% RH, a detection range of 0-100% relative humidity, and a response time of less than 8 seconds. It is used to monitor the relative humidity changes of the storage environment in real time.

[0118] 3) pH sensor module: It uses a digital pH sensor with an accuracy of ±0.01pH, a measurement range of 3-9pH, and an automatic temperature compensation function. It is used to monitor the pH value changes of microbial fertilizers.

[0119] See also Figure 2 These sensor modules are connected to the microprocessor through signal conditioning circuits. The microprocessor adopts ARM architecture, has a main frequency of no less than 1GHz, and has a built-in RAM of no less than 512MB. The microprocessor pre-processes the collected data, including:

[0120] Outlier detection and filtering;

[0121] Data smoothing;

[0122] Standardization of data formats;

[0123] The specific method for the laboratory testing unit to collect bacterial activity data is as follows:

[0124] 1) Count colonies using an automatic colony counting system;

[0125] 2) Morphological observation using a digital microscope;

[0126] 3) Measure bacterial density using a spectrophotometer;

[0127] S1.2 Data upload stage:

[0128] The edge computing node uploads the collected microbial data to the cloud computing server.

[0129] S1.2 Specific implementation of the data upload phase;

[0130] The edge computing node uploads data through the industrial Ethernet communication module, including:

[0131] 1) Data packaging: Pack the preprocessed data according to the predefined protocol format;

[0132] 2) Data encryption: AES-256 encryption algorithm is used to ensure data transmission security;

[0133] 3) Data upload: Send data packets to the cloud computing server through the MQTT protocol;

[0134] 4) Transmission confirmation: implement the confirmation mechanism for data transmission to ensure data integrity;

[0135] S1.3 Model building phase:

[0136] The cloud computing server builds a formula optimization model based on temperature, humidity, pH parameters and bacterial activity data.

[0137] Specific implementation of the model building phase:

[0138] Cloud computing servers use a distributed architecture, including;

[0139] 1) Data processing node: equipped with a high-performance processor with a main frequency of no less than 2.5 GHz;

[0140] 2) Storage node: using a distributed database with a storage capacity of no less than 10TB;

[0141] 3) Computing nodes: Configure GPU accelerator cards for model training and optimization calculations;

[0142] The process of building a recipe optimization model includes:

[0143] 1) Data preprocessing: cleaning and standardizing the collected data;

[0144] 2) Feature engineering: extract key features and construct feature vectors;

[0145] 3) Model training: Use machine learning algorithms to train the model;

[0146] 4) Model validation: Evaluate model performance through cross-validation;

[0147] 5) Model optimization: optimize model parameters according to the verification results;

[0148] In one of the embodiments, during the data preprocessing stage, the system first performs outlier detection on the collected environmental parameters (temperature, humidity, pH value) and bacterial activity data, and removes data points that are obviously deviated. Then the data is normalized to unify the parameters of different dimensions to the same scale. For temperature data, it is standardized to the range of 20-40°C; humidity data is standardized to the range of 0-100%; pH value is standardized to the range of 3-9. At the same time, the colony count data is logarithmically transformed to make it more consistent with the normal distribution characteristics.

[0149] The feature engineering phase focuses on extracting features that have a significant impact on bacterial activity. Principal component analysis found that temperature, pH value and storage time are the dominant factors, with a cumulative contribution rate of more than 85%. Among them, the contribution rate of temperature is 45%, the contribution rate of pH value is 25%, and the contribution rate of storage time is 15%. Based on this, the feature vector is constructed, and the original 7-dimensional input space (including temperature, humidity, pH value, storage time, protective agent concentration, preservative concentration, NPK ratio type) is simplified to a feature space with these three dominant factors as the core.

[0150] The model training adopts a hierarchical modeling strategy. The first layer deals with the effects of temperature, humidity and pH value, the second layer describes the time evolution law, and the third layer deals with the effects of formulation parameters. The exponential function f1(T) = α × exp(-β(T-25)) is used to model the temperature effect, and the Gaussian function f2(P) = exp(-(P-P0) is used to model the pH effect. 2 / σ 2) is used to describe the time decay, and the modified exponential function f3(t) = exp(-γt) is used. The parameters of these functions are optimized through machine learning algorithms.

[0151] The cross-validation method was used to evaluate the performance during the model validation phase. Under normal temperature conditions, the average error between the model prediction and the measured value was controlled within 8%; under high temperature conditions, the error increased slightly but was still controlled within 12%. For long-term storage, within a storage period of 4 months, more than 90% of the predicted values ​​under normal temperature conditions were within 10% of the measured values, and more than 80% of the predicted values ​​under high temperature conditions were within 15% of the measured values.

[0152] Finally, in the model optimization stage, the model parameters are dynamically adjusted according to the verification results. The gradient descent method is used for optimization, and the root mean square of the prediction error is used as the optimization objective function. The β value of the temperature influence function is optimized to the range of 0.15-0.18, the optimal pH value of the pH response function is determined to be in the range of 6.8-7.2, and the pH sensitivity parameter σ is about 1.5. The time attenuation coefficient γ is about 0.005 at room temperature and increases to 0.008 at high temperature. Through repeated iterative optimization, a set of parameter values ​​that can better balance the prediction accuracy and model complexity are finally determined.

[0153] S1.4 Solution output stage:

[0154] The cloud computing server calculates and outputs the optimal ratio scheme based on the formula optimization model, where the protective agent concentration range is 0-5‰ and the preservative concentration range is 0-2‰, and controls the edge computing nodes to monitor and verify the bacterial activity stability of the optimal ratio scheme under different storage conditions.

[0155] S1.4 Specific implementation of the solution output stage;

[0156] The process of calculating the optimal ratio scheme by the cloud computing server according to the optimization model includes:

[0157] 1) Parameter range setting;

[0158] Protectant concentration range: 0-5‰;

[0159] Preservative concentration range: 0-2‰;

[0160] 2) Generation of matching scheme;

[0161] Establish the objective function: maximize the bacterial viability retention rate;

[0162] Constraint setting: Consider cost and process limitations;

[0163] Optimization algorithm solution: Genetic algorithm is used to find the optimal solution;

[0164] 3) Verification control;

[0165] Generate verification scheme;

[0166] Issue control instructions;

[0167] monitoring implementation;

[0168] S1.5 Dynamic adjustment stage:

[0169] The cloud computing server receives the bacterial activity detection data regularly uploaded by the edge computing nodes, performs stability analysis, and dynamically adjusts the ratio parameters of the protective agent and the preservative according to the analysis results.

[0170] S1.5 Specific implementation of the dynamic adjustment phase;

[0171] The specific steps for cloud computing servers to achieve dynamic adjustment are:

[0172] 1) Data collection;

[0173] Receive regularly uploaded bacterial activity test data;

[0174] Record changes in environmental parameters;

[0175] Statistics on recipe execution effect;

[0176] 2) Stability analysis;

[0177] Calculate the trend of bacterial activity;

[0178] Analyze influencing factors;

[0179] Assess the need for adjustments;

[0180] 3) Parameter adjustment;

[0181] Calculate the adjustment amount based on the analysis results;

[0182] Generate new ratio parameters;

[0183] Verify the effect of the adjustment.

[0184] In one embodiment,

[0185] S2: Specific implementation of microbial data collection

[0186] like Figure 3 First, before collecting microbial data, the laboratory testing unit needs to be prepared. The laboratory testing unit includes an automatic colony counting system, a digital microscope with image analysis function, an environmental test chamber with controlled temperature and humidity, and a sterile operating table. The laboratory testing unit needs to be disinfected and calibrated according to standard laboratory operating procedures to ensure the accuracy of the test results.

[0187] In terms of the colony count determination of the basic formula, the two typical NPK formulas of 20-20-20+TE and 12-6-40+TE are preferred as the basic formulas. These two formulas represent water-soluble fertilizers with different nutrient ratios. The 20-20-20+TE formula has a balanced content of nitrogen, phosphorus and potassium, while the 12-6-40+TE formula is more potassium-heavy. For each formula, the standard plate count method is used for colony count. In the specific operation, the sample is first diluted in series, and the appropriate dilution is selected. The diluted sample is inoculated on the culture medium plate under sterile conditions, and the sample is placed in a constant temperature incubator at 28°C for 48 hours before counting.

[0188] In the strain configuration stage, two functional bacteria, Bacillus subtilis and Bacillus amyloliquefaciens, were selected. Both strains are commonly used soil improvement bacteria with strong environmental adaptability. The two strains were mixed in a 1:1 ratio, and the initial amount of added bacteria was controlled at 4.00E+08CFU / g. The activation and expansion process of the strains requires strict control of temperature and culture time to ensure the activity of the strains. Aseptic operation is used in the mixing process to avoid cross contamination.

[0189] In order to accurately grasp the stability changes of the product, a systematic monitoring mechanism has been established. The colony count and pH value are monitored regularly every month. When counting colonies, an automatic colony counting system is used, which is equipped with a high-definition camera and intelligent recognition software to complete the counting work quickly and accurately. The pH value is determined using a high-precision digital pH meter, which needs to be calibrated before each measurement. In addition, the ambient temperature and relative humidity data during the test need to be recorded to provide complete reference information for subsequent analysis.

[0190] In terms of survival rate analysis, the focus is on the survival of bacteria under different pH conditions. By adjusting the pH value of the formula, multiple test points are set in the range of 3-9, the number of colonies is measured simultaneously under each pH condition, and the survival rate relative to the initial bacterial count is calculated. This process requires special attention to the selection of pH regulators to avoid introducing ions that have a significant effect on the activity of the strain. Experimental data show that the survival rate of bacteria is generally high in the pH range of 4.5-8.0, while the survival rate will be significantly reduced when the pH is lower than 4.5 or higher than 8.0.

[0191] Finally, in terms of correlation recording, a complete data recording system has been established. The recorded content includes basic data such as ambient temperature, relative humidity, pH value, colony count, etc., as well as records of phenomena such as changes in sample appearance and precipitation. All data are managed using special experimental recording software to ensure data traceability. Through statistical analysis of these data, the correlation between environmental condition parameters and bacterial activity can be revealed, providing an important basis for subsequent formula optimization.

[0192] Practice has shown that this systematic microbial data collection method not only ensures the accuracy and completeness of the data, but also can deeply reflect the stability characteristics of the product under different conditions. Through standardized operating procedures and advanced testing equipment, the efficiency and accuracy of data collection have been significantly improved. These detailed basic data provide reliable support for the subsequent formulation optimization model construction.

[0193] In order to more intuitively illustrate the technical effect of the present invention, a set of typical experimental data before and after optimization is given below. The experiment selected 20-20-20+TE formula and 12-6-40+TE formula as the basic formula, and carried out stability tests for 4 months under different storage conditions. The specific data are as follows:

[0194] Table 1 Comparative data of bacterial activity of 20-20-20+TE formula under different storage conditions

[0195] Storage conditions Testing time (month) Bacterial activity before optimization (CFU / g) Optimized bacterial activity (CFU / g) pH Range Normal temperature and humidity (25℃, 45%) 0 4.00E+08 4.00E+08 4.5-7.2 1 3.25E+08 3.85E+08 4.5-7.2 2 2.85E+08 3.62E+08 4.6-7.1 3 2.32E+08 3.45E+08 4.6-7.1 4 2.05E+08 3.28E+08 4.7-7.0

[0196] Table 2 Comparative data of bacterial activity of 12-6-40+TE formula under high temperature and high humidity conditions

[0197]

[0198] From the above experimental data, it can be seen that the optimized formula shows obvious advantages under various storage conditions:

[0199] Under normal temperature and humidity conditions, the bacterial activity retention rate of the optimized formula reached 82% after 4 months, while it was only 51% before optimization.

[0200] Under extreme conditions of high temperature and high humidity, the bacterial activity of the optimized formula could still be maintained above 2.45E+08CFU / g for 4 months, while the bacterial activity of the formula before optimization had dropped to 0.98E+08CFU / g, which was significantly lower than the target value of 2.00E+08CFU / g.

[0201] The stability of pH value has also been significantly improved, and the pH fluctuation range of the optimized formula has been significantly narrowed, which plays an important role in maintaining the stability of bacterial activity.

[0202] The specific implementation of S2 microbial data collection includes:

[0203] S2.1 Colony count determination of basic formula:

[0204] The standard plate count method was used to count the colonies of the two basic formulas, 20-20-20+TE and 12-6-40+TE.

[0205] S2.2 Bacteria configuration:

[0206] In each basic formula, Bacillus subtilis and Bacillus amyloliquefaciens were mixed in a ratio of 1:1, and the initial amount of bacteria added was 4.00E+08 CFU / g.

[0207] S2.3 Monitoring frequency:

[0208] Monitor colony counts and pH changes monthly.

[0209] S2.4 Survival analysis:

[0210] The bacterial survival rate was analyzed under different pH conditions.

[0211] S2.5 Related records:

[0212] Correlation data between environmental condition parameters and bacterial activity were recorded.

[0213] Exemplarily, the standard plate count method was used to conduct colony counting experiments on different fertilizer formulations. The experiment selected 10 commonly used raw materials, 4 common macronutrient water-soluble fertilizers and 2 fertilizers with different pH formulas, and a total of 16 samples were tested. The first viable bacteria count showed that under the condition of an addition amount of 4.00E+08CFU / g, except for the urea phosphate raw material, the viable bacteria counts of other samples basically met the standards. It is worth noting that the pH value has a significant effect on the number of viable bacteria. The viable bacteria counts of samples with a pH value below 4.5 are significantly lower, such as Figure 7 shown.

[0214] Secondly, the survival rate of bacteria under different storage conditions was analyzed. The experiment in the report divided the samples into four groups: normal temperature and humidity, normal temperature and high humidity, high temperature and humidity, and high temperature and high humidity. The experimental results showed that under normal temperature conditions, such as Figure 8a and Figure 8b , except for the acidic fertilizer sample, all other fertilizer samples reached the standard of 2.00E+08CFU / g in the fourth month.

[0215] Finally, the correlation analysis between environmental factors and bacterial activity was performed. According to the relevant data, under high temperature and high humidity conditions, the number of live bacteria in all fertilizer samples, except for the alkaline formula, decreased sharply. In particular, the 12-6-40 formula can still maintain a high number of live bacteria under high temperature and normal humidity conditions, but the number of live bacteria drops sharply under high temperature and high humidity conditions, as shown in Figure 8c.

[0216] S3: Construction of formulation optimization model

[0217] like Figure 4In the process of constructing the formula optimization model, it is first necessary to establish a high-dimensional function approximation model. The model takes temperature, humidity, pH value, storage time, protective agent concentration and preservative concentration as input variables to construct a high-dimensional feature space. Each variable has its specific range of values: temperature is between 20-40°C, relative humidity is between 0-100%, pH value is between 3-9, storage time is recorded in days, protective agent concentration is in the range of 0-5‰, and preservative concentration is in the range of 0-2‰. These variables together constitute a six-dimensional feature space.

[0218] In order to improve the computational efficiency of the model, it is necessary to identify effective dimensions that have a significant impact on bacterial activity from these input variables. Through the analysis of experimental data, it is found that temperature, pH value and storage time are the main factors affecting bacterial activity. Among them, temperature has the most direct effect on bacterial activity. Under high temperature conditions of 40°C, bacterial activity will decrease rapidly; the influence of pH value is second. In the range of 4.5-8.0, bacterial activity remains relatively stable; storage time shows a gradual impact, and bacterial activity shows a slow downward trend over time.

[0219] Based on the above findings, the embodiment of the present application simplifies the original high-dimensional model. By introducing weight coefficients, the role of the main influencing factors is highlighted, and the secondary factors are used as correction items. This simplification not only greatly reduces the complexity of the model, but also retains the essential characteristics of the problem. Specifically, the original six-dimensional model is simplified to a three-dimensional model dominated by temperature, pH value and time, and other factors are corrected by influencing coefficients.

[0220] In order to verify the accuracy of the simplified model, a series of verification experiments were designed in the embodiments of the present application. Experimental data were collected under different conditions, and the actual observed values ​​were compared with the model prediction values. The verification results show that the prediction error of the simplified model is within an acceptable range, and the prediction accuracy under normal temperature and humidity conditions is more than 90%, and even under extreme conditions, the prediction accuracy can be maintained at more than 80%.

[0221] Finally, the model parameters are optimized based on the verification results. Specifically, the weight coefficients are adjusted, the influencing factors are corrected, and the threshold parameters are updated. Through iterative optimization, the prediction accuracy of the model is continuously improved. The optimization process uses the gradient descent method, and the root mean square of the prediction error is used as the optimization objective function. After multiple rounds of optimization, the overall performance of the model has been significantly improved.

[0222] In practical applications, this simplified model shows good applicability. It can not only accurately predict the trend of bacterial activity changes under different conditions, but also has high calculation efficiency and is easy to implement in engineering. Especially for scenarios where rapid adjustment of formulas is required, the model can provide timely optimization suggestions and provide strong support for production practice.

[0223] S3.1 Construction of high-dimensional function approximation model:

[0224] Temperature, humidity, pH value, storage time, protective agent concentration and preservative concentration were used as input variables to construct a high-dimensional feature space.

[0225] First, construct a set of input variables. The input variables include environmental condition parameters (temperature T, relative humidity H, pH value P, storage time t) and formula parameters (protectant concentration C1, preservative concentration C2, NPK ratio type N). Among them, the experimental range of temperature is 20-40°C, the relative humidity range is 0-100%, the pH value range is 3-9, the storage time range is 0-120 days, the protective agent concentration range is 0-5‰, the preservative concentration range is 0-2‰, and the NPK ratio type is a discrete variable. In this way, the embodiment of the present application actually faces a 7-dimensional input space.

[0226] When dealing with high-dimensional feature space, a hierarchical modeling strategy is adopted. The first layer is the environmental parameter layer, which mainly deals with the effects of temperature, humidity and pH value; the second layer is the time evolution layer, which describes the change law over time under different environmental conditions; the third layer is the formula parameter layer, which deals with the influence of the ratio of protective agents and preservatives. This hierarchical structure makes complex high-dimensional problems easier to handle.

[0227] Based on the experimental data, the functional relationship between bacterial activity A and each input variable is established. The initial model can be expressed as:

[0228] A=f(T,H,P,t,C1,C2,N).

[0229] By analyzing the distribution characteristics of experimental data, it is found that the function can be simplified to:

[0230] A≈g(T,P,t)×h(H,C1,C2)×k(N),

[0231] Among them, g(T,P,t) represents the role of the main influencing factors, h(H,C1,C2) represents the regulatory role of the secondary factors, and k(N) represents the influence coefficient of the matching type.

[0232] S3.2 Effective dimension identification:

[0233] Identify effective dimensions that have a significant impact on bacterial activity from input variables.

[0234] In order to accurately identify the effective dimensions, we first conducted a sensitivity analysis of each variable. When other conditions remained unchanged, we changed each variable separately and recorded the changes in bacterial activity. The analysis results showed that for every 5°C increase in temperature, bacterial activity decreased by an average of 15-20%; for every 0.5 unit deviation of the pH value from the optimal range, bacterial activity decreased by an average of 10-15%; and the change in relative humidity had a relatively small effect on bacterial activity, with a 20% change in humidity resulting in a change in bacterial activity of no more than 5%.

[0235] The dominant factors were further confirmed by principal component analysis. The results showed that the cumulative contribution of the three dimensions of temperature, pH value and storage time exceeded 85%, which can be used as the main dimensions of the model. Among them, the contribution rate of temperature was the highest, reaching 45%; pH value was second, with a contribution rate of 25%; and the contribution rate of storage time was 15%. This finding provides an important basis for model simplification.

[0236] S3.3 Model simplification:

[0237] Reduce model complexity based on effective dimensions and build a simplified model.

[0238] Based on the results of effective dimension identification, the original 7-dimensional model was simplified to a 3-dimensional dominant model. The simplified model focuses on describing the influence of three main factors: temperature, pH value, and time, and other factors are treated as correction terms. This simplification not only greatly reduces the computational complexity, but also maintains the core predictive ability of the model.

[0239] In the simplified model, the temperature effect is described by an exponential function: f1(T) = α × exp(-β(T-25)), where α is the initial bacterial activity coefficient and β is the temperature sensitivity coefficient. The pH effect is described by a Gaussian function: f2(P) = exp(-(P-P0) 2 / σ 2 ), where P0 is the optimal pH value and σ is the pH sensitivity parameter. The time decay uses a modified exponential decay function: f3(t)=exp(-γt), where γ is the time decay coefficient.

[0240] S3.4 Model Validation:

[0241] The prediction accuracy of the simplified model was verified by experimental data.

[0242] In the model verification phase, a systematic verification experiment plan was designed. First, at room temperature (25°C), the formula with different pH values ​​was used for verification. The results showed that the average error between the model prediction value and the measured value was within 8%. Then, the model was verified under high temperature conditions (40°C). Although the error increased slightly, it was still controlled within 12%. This shows that the simplified model has good predictive ability.

[0243] The verification was especially carried out for long-term storage. During the 4-month storage period, samples were taken and tested once a month, and the measured bacterial activity data was compared with the model prediction value. The results showed that under normal temperature conditions, the error between the predicted value and the measured value was within 10% for more than 90%; even under high temperature conditions, the error between the predicted value and the measured value was controlled within 15% for more than 80%.

[0244] S3.5 parameter optimization:

[0245] Optimize and simplify the model parameters based on the verification results.

[0246] According to the verification results, the iterative optimization method was used to optimize the model parameters. First, the α and β parameters in the temperature influence function were determined by the least squares method. Analysis found that when the β value was in the range of 0.15-0.18, the temperature response of the model was most accurate. Secondly, the parameters of the pH response function were optimized, and the optimal pH value P0 was determined to be in the range of 6.8-7.2, and the pH sensitivity parameter σ was approximately 1.5. Finally, the time attenuation coefficient γ was optimized based on long-term storage data, and it was found that the γ value was correlated with temperature, which was approximately 0.005 at room temperature and increased to 0.008 at high temperature.

[0247] The parameter optimization process uses the gradient descent method, with the root mean square of the prediction error as the optimization objective function. The optimization process is divided into multiple iterations, and each cycle is verified and adjusted using new experimental data. Through repeated iterations, a set of parameter values ​​that can better balance prediction accuracy and model complexity is finally determined.

[0248] The model after parameter optimization showed good prediction performance and practicality. In practical applications, the model can not only accurately predict the trend of bacterial activity under different conditions, but also has high computational efficiency and is easy to implement in engineering. In particular, for scenarios that require rapid adjustment of formulations, the model can provide timely optimization suggestions.

[0249] For example,

[0250] In the system of optimizing the ratio of microbial fertilizer protective agent and preservative, the input variables include:

[0251] 1) Environmental condition parameters:

[0252] Temperature (T): experimental range 20-40℃;

[0253] Relative humidity (H): 0-100%;

[0254] pH value (P): 3-9;

[0255] Storage time (t): 0-120 days;

[0256] 2) Recipe parameters:

[0257] Protectant concentration (C1): 0-5‰;

[0258] Preservative concentration (C2): 0-2‰;

[0259] NPK ratio type (N): categorical variable;

[0260] In this way, we are actually facing a 7-dimensional input space. However, by analyzing the experimental data, we found that there are significant opportunities for dimensional simplification. For example, when the temperature is fixed at 40°C, even if the humidity conditions are adjusted, the change in bacterial activity is mainly affected by the pH value. Specifically, in the alkaline formula (pH = 7.07), the bacterial activity can be maintained at 4.05E+08CFU / g, while the bacterial activity in the acidic formula (pH = 3.64) is significantly reduced.

[0261] Based on this discovery, a constructive approximation method for high-dimensional functions with smaller effective dimensions is adopted. The specific implementation steps are as follows:

[0262] 1. Dimension importance analysis

[0263] Through statistical analysis of experimental data, the dominant dimensions were determined. From the experimental data, the three dimensions that have the most significant impact on bacterial activity are:

[0264] Temperature (T): Experiments show that bacterial activity decreases significantly at 40°C.

[0265] pH value (P): The bacterial activity is best when the pH is in the range of 4.5-8;

[0266] Storage time (t): The effect of time on bacterial activity shows a gradual attenuation characteristic;

[0267] 2. Build a simplified model

[0268] The bacterial activity A is expressed as a function of each parameter:

[0269] A=f(T,H,P,t,C1,C2,N)

[0270] By analyzing the experimental data, the function can be simplified to:

[0271] A≈g(T,P,t)×h(H,C1,C2)×k(N)

[0272] in:

[0273] g(T,P,t) represents the effect of the main influencing factors;

[0274] h(H,C1,C2) represents the moderating effect of the secondary factors;

[0275] k(N) represents the influence coefficient of the ratio type;

[0276] 3. Quantitative Relationship Modeling

[0277] Taking the three-dimensional space of temperature-pH-time as an example, a specific mathematical relationship is established based on experimental data:

[0278] g(T,P,t)=α×exp(-β(T-25))×f(P)×exp(-γt)

[0279] in:

[0280] α is the initial bacterial activity coefficient;

[0281] β is the temperature sensitivity coefficient;

[0282] f(P) is the pH response function, which takes its maximum value near pH = 7;

[0283] γ is the time attenuation coefficient;

[0284] Verify the experimental data:

[0285] Under normal temperature (25°C), the alkaline formula with pH = 7.07 maintained bacterial activity above 3.31E+08 CFU / g for 4 months;

[0286] Under high temperature (40°C), the bacterial activity of the same formula was reduced to 3.01E+08CFU / g;

[0287] The bacterial activity of the acidic formula with pH=3.64 was significantly lower than 2.00E+08CFU / g under various conditions;

[0288] This simplified model not only retains the core features of the problem, but also greatly reduces the computational complexity. The good agreement between the model prediction results and the experimental data verifies the effectiveness of this dimensional simplification method. This is the specific embodiment of the invention point of "constructive approximation of high-dimensional functions with small effective dimensions".

[0289] This step analyzes experimental data, identifies key dimensions that affect microbial activity, and constructs a simplified but effective mathematical model. This approach not only reduces computational complexity, but also maintains the model's predictive accuracy, providing a reliable theoretical basis for subsequent formulation optimization.

[0290] S4: Determination of dominant dimensions and model construction

[0291] When determining the dominant dimension, we first conduct an in-depth analysis of the impact of temperature. Through a large amount of experimental data, we found that the impact of temperature on bacterial activity presents obvious nonlinear characteristics. In the range of 20-30℃, bacterial activity remains relatively stable; when the temperature rises above 35℃, bacterial activity begins to decline rapidly; at 40℃, bacterial activity decreases significantly. Based on this feature, a temperature influence function is established, and an exponential function is used to describe the relationship between temperature and bacterial activity.

[0292] pH value is the second dominant dimension that affects bacterial activity. Through systematic pH scanning experiments, it was found that there is a clear correlation between pH value and bacterial activity. In the pH range of 4.5-8.0, bacterial activity is maintained at a high level; when the pH value is lower than 4.5 or higher than 8.0, bacterial activity will drop rapidly. This relationship can be described by a Gaussian function, with its central value set around pH = 6.5, which is consistent with the physiological characteristics of microorganisms.

[0293] A progressive decay model was established for the third dominant dimension, storage time. Experimental data show that bacterial activity exhibits typical exponential decay characteristics as storage time increases, but the decay rate slows down over time. This trend can be described by a modified exponential decay function, in which the decay coefficient is related to temperature and pH. For example, at room temperature (25°C), when the pH value is maintained in the range of 6.5-7.5, the bacterial activity decay does not exceed 20% during a storage period of 4 months.

[0294] The three dominant dimensions are integrated together to construct a comprehensive simplified model. The model adopts a hierarchical structure, with the first layer being the temperature influence function, the second layer being the pH response function, and the third layer being the time decay function. The three functions are interrelated through coupling coefficients to form a complete mathematical description. The general form of the model can be expressed as:

[0295] A(t,T,pH)=A0·f(T)·g(pH)·h(t)

[0296] Among them, A0 is the initial bacterial activity value, f(T) is the temperature influence function, g(pH) is the pH response function, and h(t) is the time decay function.

[0297] Based on a large amount of experimental data, the least squares method was used to determine the parameters in the model. First, the parameters of the temperature effect function f(T) were determined through a constant temperature experiment; then, under the optimal temperature condition, the parameters of the pH response function g(pH) were determined through a pH gradient experiment; finally, the parameters of the time decay function h(t) were determined through a long-term storage experiment.

[0298] S4.1 Temperature influence analysis:

[0299] Determine temperature as the first dominant dimension and establish the temperature influence function.

[0300] Through a systematic temperature gradient experiment, the embodiment of the present application has studied in detail the influence of temperature on the activity of microorganisms. In the range of 20-25°C, the activity remains relatively stable, and the decay rate does not exceed 5% / month. When the temperature rises to 30-35°C, the activity begins to decline significantly, and the monthly decay rate increases to 8-12%. In the high temperature range of 35-40°C, the activity drops rapidly, and the monthly decay rate reaches 15-20%. This nonlinear temperature response characteristic requires the embodiment of the present application to use a more complex mathematical model to describe it.

[0301] Based on experimental data, the present invention establishes an improved exponential temperature influence function: f(T) = α·exp(-β(T-T0) 2 ), where T0 is the optimum temperature (about 25°C), α is the initial activity coefficient, and β is the temperature sensitivity coefficient. Through data fitting, it was found that when β is in the range of 0.015-0.018, the model can well describe the effect of temperature on bacterial activity. Especially in high temperature areas, the error between the model prediction value and the measured value is controlled within 10%.

[0302] S4.2 pH correlation analysis:

[0303] The correlation between pH value and bacterial activity was analyzed, and pH value was determined as the second dominant dimension.

[0304] The effect of pH value on bacterial activity shows an obvious "bell-shaped" characteristic. Through pH gradient experiments, the present application examples found that below pH 4.5, bacterial activity dropped sharply, with a survival rate of less than 30%; in the weakly acidic range of pH 4.5-6.5, bacterial activity gradually increased, with a survival rate of 60-80%; in the neutral range of pH 6.5-7.5, bacterial activity reached the highest, with a survival rate of more than 90%; and in the alkaline range above pH 7.5, bacterial activity began to decrease again, but the rate of decrease was slower than that under acidic conditions.

[0305] In view of this feature, the present invention designs a double Gaussian function to describe the pH response: g(pH)=exp(-(pH-pH0) 2 / σ1 2 )+k·exp(-(pH-pH1) 2 / σ2 2 ). Where pH0 and pH1 are the two optimal pH values ​​(about 6.8 and 7.2), σ1 and σ2 are the corresponding pH sensitivity parameters, and k is the weight coefficient. This function can well fit the change trend of bacterial activity observed under different pH conditions.

[0306] S4.3 Time impact model:

[0307] Establish an asymptotic decay model of storage time as the third dominant dimension.

[0308] The effect of storage time on viable bacteria shows typical asymptotic decay characteristics, but this decay is not a simple exponential decay. It is observed in the embodiments of the present application that in the initial stage of storage (0 - 30 days), the viable bacteria decay rapidly, with a monthly decay rate of about 15 - 20%; then it enters a relatively stable period (30 - 90 days), and the monthly decay rate drops to 5 - 8%; in the long-term storage stage (more than 90 days), the decay rate further slows down, and the monthly decay rate remains at 3 - 5%.

[0309] To accurately describe this non-uniform decay characteristic, the embodiments of the present application propose a piecewise exponential decay model: h(t) = a1·exp(-γ1t), t ≤ 30 days; h(t) = a2·exp(-γ2t), 30 days < t ≤ 90 days; h(t) = a3·exp(-γ3t), t > 90 days. Where a1, a2, and a3 are the initial activity coefficients of each stage, and γ1, γ2, and γ3 are the corresponding decay coefficients. Through parameter optimization, the goodness of fit R of the model prediction value and the measured data 2 reaches above 0.95.

[0310] S4.4 Model integration:

[0311] Integrate the three dominant dimensions to construct a simplified model.

[0312] When integrating the three dimensions of temperature, pH, and time, the coupling effect between them needs to be considered. Experiments find that there is a significant interaction between temperature and pH: under high-temperature conditions, the influence of pH is more sensitive; while within the optimal pH range, the ability to tolerate high temperature is also enhanced accordingly. The time decay rate is also affected by temperature and pH, showing that the decay accelerates under high-temperature or non-optimal pH conditions.

[0313] Based on these observations, the embodiments of the present application establish a comprehensive model: F(T, pH, t) = f(T)·g(pH)·h(t) + λ1·f(T)·g(pH) + λ2·g(pH)·h(t) + λ3·f(T)·h(t). Where λ1, λ2, and λ3 are coupling term coefficients used to describe the interaction between each dimension. This model not only considers the independent influence of each factor but also includes the interaction between them.

[0314] S4.5 Parameter calculation:

[0315] Calculate the parameters of the simplified model based on experimental data.

[0316] The parameter calculation adopts a step-by-step optimization strategy. First, the parameters of the temperature function, pH function and time function are optimized respectively using single-factor experimental data. Then, the coupling term coefficient is optimized using orthogonal experimental data. Finally, the complete experimental data set is used for overall optimization, and an improved genetic algorithm is used to find the optimal parameter combination.

[0317] During the optimization process, the present embodiment found that some parameters were correlated, especially the relationship between the temperature sensitivity coefficient β and the pH sensitivity parameter σ. By introducing the constraints: 0.015≤β≤0.018, 1.2≤σ≤1.8, not only the physical significance of the model is ensured, but also the optimization efficiency is improved. The prediction accuracy of the final model under various conditions reached more than 85%, meeting the needs of engineering applications.

[0318] S5: Calculation process of the optimal ratio solution

[0319] In practical applications, the calculation of the ratio scheme first requires the systematic classification of the existing bacterial activity data. The embodiment of the present application divides the data into multiple intervals according to the pH range: strong acidity (pH <4.5), weak acidity (4.5 ≤ pH <6.5), neutrality (6.5 ≤ pH <7.5) and alkalinity (pH ≥ 7.5). In each interval, the data is further grouped according to temperature conditions and storage time to form a structured data set.

[0320] Based on the classified data set, a two-dimensional distribution function is generated on the pH-bacterial activity plane. This distribution function reflects the distribution characteristics of bacterial activity at different pH values. In order to deal with the discreteness of the data, the kernel density estimation method is used to smooth the distribution function. By selecting an appropriate bandwidth parameter, the main characteristics of the data are maintained and the overfitting problem is avoided.

[0321] The smoothed distribution function is projected to obtain projection data at different angles. These projection data are used for subsequent Radon transform. The purpose of Radon transform is to convert a two-dimensional distribution into a series of one-dimensional projections that contain complete information about the original distribution.

[0322] In order to deal with the uncertainty caused by the limited amount of experimental data, regularization processing is introduced. The selection of regularization terms is based on two principles: one is to be able to effectively suppress the influence of noise, and the other is to maintain the main characteristics of the data. The optimal regularization parameter is determined by the cross-validation method so that the model will neither overfit nor underfit.

[0323] S5.1 Data classification:

[0324] The bacterial activity data under different formulation conditions were classified to generate bacterial activity data sets within different pH ranges.

[0325] First, a systematic classification is performed based on the bacterial activity data obtained from the experiment. The embodiment of the present application divides the data into four main intervals according to the pH value: a strongly acidic interval (pH <4.5), a weakly acidic interval (4.5 ≤ pH <6.5), a neutral interval (6.5 ≤ pH ≤ 7.5) and an alkaline interval (pH> 7.5). In each pH interval, sub-intervals are further divided according to temperature conditions: a normal temperature interval (20-30 ° C) and a high temperature interval (30-40 ° C). The data of each sub-interval is further divided into three stages according to the storage time: the initial stage (0-30 days), the middle stage (31-90 days) and the late stage (91-120 days).

[0326] During the data sorting process, special attention was paid to the handling of outliers. The 3σ principle was used to screen out abnormal data, and professional knowledge was combined to ensure that the removal of outliers would not affect the representativeness of the data. After screening and sorting, a structured data set containing more than 1,000 valid data points was formed, providing a reliable data basis for subsequent analysis.

[0327] S5.2 Data projection:

[0328] Based on the bacterial activity data set, a two-dimensional distribution function was generated on the pH-bacterial activity plane and the data was projected to obtain the projection data.

[0329] Based on the sorted data set, a two-dimensional distribution function was constructed on the pH-bacterial activity plane. The discrete data points were first smoothed using the kernel density estimation method, and the Gaussian kernel function was selected as the basic kernel function. The kernel width was optimized and determined by the cross-validation method. Under different temperature and storage time conditions, the corresponding two-dimensional distribution functions were generated.

[0330] The data projection process adopts an equi-angle projection strategy, and performs a projection every 5 degrees within the range of 0-180 degrees, obtaining data in 36 projection directions. The data in each projection direction contains part of the original distribution information. In order to improve the accuracy of the projection, the interpolation algorithm is used to encrypt the data during the projection process so that the projection data can better reflect the characteristics of the original distribution.

[0331] S5.3 Radon transform:

[0332] Perform Radon transform on the projection data to obtain the transformed distribution function.

[0333] The purpose of the Radon transform is to transform a two-dimensional distribution into a series of one-dimensional projections. For each projection direction θ, the Radon transform is calculated:

[0334] R[f](θ,s)=∫∫f(x,y)δ(s xcosθysinθ)dxdy.

[0335] Where f(x,y) is the two-dimensional distribution function on the pH-bacterial activity plane, and δ is the Dirac function. In order to improve the computational efficiency, the fast Radon transform algorithm is used to implement the integral operation through discretization.

[0336] In actual calculations, the present application embodiment found that the information redundancy of some projection directions was high. By analyzing the correlation of the projection data, representative projection directions were screened out, and finally the data of 12 key projection directions were retained. This screening not only reduces the amount of calculation, but also improves the efficiency of subsequent processing.

[0337] S5.4 Regularization processing:

[0338] A regularization term is introduced into the transformed distribution function to deal with the uncertainty caused by the limited number of samples and obtain a regularized distribution estimate.

[0339] S5.4 Specific implementation of regularization processing

[0340] In order to deal with the uncertainty caused by the limited number of samples, the Tikhonov regularization method is introduced. The regularization objective function is set as: E(f) = ||Rf g|| 2 +λΦ(f), where Rf represents the Radon transform of f, g represents the observed data, λ is the regularization parameter, and Φ(f) is the regularization term. The regularization parameter λ is selected using the L-curve method to find the best balance between maintaining data features and suppressing noise.

[0341] Under different temperature and pH conditions, the regularization parameter needs to be adjusted dynamically. For example, in the pH-sensitive region (pH < 4.5 or pH > 8), the regularization parameter needs to be increased due to large data fluctuations; while in the pH-stable region, the regularization parameter can be appropriately reduced to retain more detailed information.

[0342] S5.5 Probability distribution modeling:

[0343] Based on regularized distribution estimation, a probability distribution model for predicting bacterial activity levels under different pH conditions was constructed.

[0344] S5.5 Construction of probability distribution model

[0345] Based on the regularized distribution estimation, a probability distribution model for predicting bacterial activity levels under different pH conditions was constructed. The model was in the form of a piecewise function: in the pH stable region (6.5≤pH≤7.5), a Gaussian distribution was used to describe the fluctuation of bacterial activity levels; in the pH transition region (4.5-6.5 and 7.5-8.5), a modified exponential distribution was used; in the pH extreme region (pH<4.5 or pH>8.5), a power law distribution was used to describe the rapid decay characteristics.

[0346] The parameters of the probability distribution model were determined by the maximum likelihood estimation method. In order to improve the predictive ability of the model, temperature and time were introduced as adjustment factors so that the model could adapt to different environmental conditions. The prediction results of the model showed that within the optimal pH range, the probability of maintaining the target level of bacterial activity exceeded 95%; in the transition area, this probability dropped to 70-85%; in the extreme area, the probability was less than 50%.

[0347] Taking actual experimental data as an example, we first preprocess the bacterial activity data under different conditions:

[0348] Observation data under normal temperature and humidity conditions:

[0349] 12-6-40 formula: 4.66E+08 CFU / g (pH=4.6);

[0350] 20-20-20 formula: 4.76E+08 CFU / g (pH=4.7);

[0351] Alkaline formula: 4.05E+08CFU / g (pH=7.07);

[0352] Observation data under high temperature and high humidity conditions:

[0353] 12-6-40 formula: 4.53E+08 CFU / g;

[0354] 20-20-20 formula: 5.24E+08 CFU / g;

[0355] Alkaline formula: 3.91E+08CFU / g;

[0356] 2. Implementation process of Radon transform

[0357] The experimental data points are projected on the pH-bacterial activity plane to obtain a series of projection data. The data distribution characteristics in the pH range of 4.5-8 can be observed. Radon transform is performed on these projection data:

[0358] R[f](θ,s)=∫∫f(x,y)δ(s xcosθysinθ)dxdy

[0359] in:

[0360] f(x,y) represents the two-dimensional distribution function of pH value and bacterial activity;

[0361] θ represents the projection angle;

[0362] s represents the projection distance;

[0363] δ represents the Dirac function;

[0364] 3. Regularization processing;

[0365] In order to deal with the uncertainty brought by finite samples, a regularization term is introduced:

[0366] E(f)=||Rf g|| 2 +λΦ(f)

[0367] in:

[0368] Rf represents the Radon transform of f;

[0369] g represents the observed data;

[0370] λ is the regularization parameter;

[0371] Φ(f) is the regularization term;

[0372] Determine the regularization parameter λ based on experimental data. For example:

[0373] Under normal temperature conditions, the pH value is relatively stable;

[0374] Under high temperature conditions, the effect of pH value is more significant;

[0375] Based on this, the regularization parameters can be adjusted dynamically.

[0376] 4. Probability distribution estimation

[0377] Based on the regularized Radon transform results, the probability distribution of maintaining the target bacterial activity level (above 2.00E+08CFU / g) under different conditions can be estimated. For example:

[0378] Probability estimation under normal temperature conditions:

[0379] pH 4.5-8 range: Maintain target bacteria survival probability > 90%

[0380] pH<4.5 range: the probability of maintaining target bacteria alive is <50%

[0381] Probability estimation under high temperature conditions:

[0382] pH 6.5-8 range: Maintain target bacteria survival probability > 80%

[0383] pH < 6.5: The probability of maintaining target bacteria alive is significantly reduced

[0384] 5. Classification decision rule formulation

[0385] According to the probability distribution estimation results, the classification decision rules for recipe optimization are established:

[0386] Under normal temperature storage conditions:

[0387] Optimum pH range: 4.5-8

[0388] Recommended formula: 12-6-40, 20-20-20 or alkaline formula

[0389] Under high temperature storage conditions:

[0390] Optimum pH range: 6.5-8

[0391] Recommended formula: mainly alkaline formula

[0392] This step effectively solves the distribution estimation problem under limited experimental data by using the maximum regularized Radon cumulative distribution transformation method. By analyzing the experimental data in the "npk+ bacteria mid-term report", not only the feasibility of this method is verified, but also the formula optimization rules with practical guiding significance are obtained. This method takes into account the uncertainty of the data and ensures the reliability of the optimization results.

[0393] S5.6 Solution Output:

[0394] Calculate and output the optimal matching plan based on the probability distribution model.

[0395] Based on the probability distribution model, a multi-objective optimization method is used to calculate the optimal ratio. The optimization objectives include: maximizing the bacterial viability retention rate, minimizing the cost, and minimizing the process complexity. By introducing the weight coefficient, the multi-objective optimization problem is transformed into a single-objective optimization problem. The optimization process uses an improved particle swarm algorithm to find the optimal solution through multiple iterations.

[0396] The calculation of the optimal ratio scheme also needs to consider the actual production constraints, including raw material supply restrictions, production process requirements and cost control targets. By setting reasonable constraints, the optimization results can be ensured to be practical. The final output ratio scheme can not only ensure product quality, but also facilitate industrial production.

[0397] S6: Construction process of probability distribution estimation model

[0398] At room temperature (25°C), the baseline probability distribution of bacterial activity levels was first established. Experimental data showed that within the optimal pH range (6.5-7.5), the probability of maintaining the target bacterial activity level (≥2.00E+08CFU / g) exceeded 90%. This probability decreased as the pH value deviated from the optimal range. When pH <4.5 or pH >8.0, the probability of maintaining the target bacterial activity level dropped to less than 30%.

[0399] For high temperature conditions (40°C), the probability distribution changes significantly. Even within the optimal pH range, the probability of maintaining the target bacterial activity level drops to about 70%. Especially under high temperature and high humidity conditions, the probability further drops to below 50%. This shows that temperature is a key factor affecting the stability of bacterial activity.

[0400] Based on these probability distribution characteristics, classification decision rules for formula optimization were established. Under normal temperature conditions, formulas with a pH value in the range of 6.5-7.5 are preferred; under high temperature conditions, special protective formulas are required and closely tracked through a monitoring system.

[0401] S6.1 Estimation under normal temperature conditions:

[0402] Estimate the probability distribution of maintaining the target bacterial activity level under normal temperature conditions (25°C).

[0403] At room temperature (25°C), a baseline probability distribution model is first established. By analyzing a large amount of experimental data, the embodiments of the present application found that in the optimal pH range (6.5-7.5), the probability distribution of maintaining bacterial activity at the target level (above 2.00E+08CFU / g) presents typical normal distribution characteristics. Specifically, when the pH value is 7.0, the probability of maintaining the target bacterial activity is the highest, reaching more than 95%; as the pH value deviates to both ends, the probability shows a symmetrical downward trend, and at pH = 6.5 and pH = 7.5, the probability can still be maintained at around 90%.

[0404] In order to accurately describe this distribution feature, the embodiment of the present application adopts an improved Gaussian distribution model:

[0405] P(A|pH)=α·exp(-(pH-7.0) 2 / 2σ 2 ).

[0406] Where A represents the bacterial activity level, α is the normalization coefficient, and σ is the distribution width parameter. Through data fitting, it was found that when σ = 0.5, the model can well describe the probability distribution observed in the experiment. In particular, in the pH range of 6.5-7.5, the deviation between the model prediction value and the experimental data does not exceed 5%.

[0407] S6.2 Estimation of high temperature conditions:

[0408] Estimate the probability distribution of maintaining the target bacterial activity level under high temperature conditions (40°C).

[0409] Under high temperature conditions (40°C), the probability distribution changes significantly. First, the overall probability level is significantly reduced, and even under the optimal pH conditions, the probability of maintaining the target bacterial activity level is reduced to 70-75%. Secondly, the shape of the probability distribution changes, no longer showing a symmetrical normal distribution feature, but showing obvious skewness. It was observed that under alkaline conditions (pH>7.0), the stability of bacterial activity is better than that under acidic conditions.

[0410] In view of this asymmetric distribution feature, the present application embodiment proposes a composite probability model: P(A|pH,T)=α·exp(-(pH-7.2) 2 / 2σ1 2 )·exp(-β(T-25)). Where T represents temperature, β is the temperature sensitivity coefficient, and σ1 is the modified distribution width parameter. This model not only takes into account the influence of pH value, but also introduces the regulatory effect of temperature factors. Experimental verification shows that under high temperature conditions, the consistency between the model prediction value and the actual observation value reaches more than 85%.

[0411] S6.3 Decision-making rule establishment:

[0412] Classification decision rules for recipe optimization are established based on probability distribution.

[0413] Based on the probability distribution model, the embodiment of the present application establishes a set of classification decision rules for formula optimization. First, the scenarios are divided into three categories according to the storage temperature conditions: room temperature storage (20-30°C), medium temperature storage (30-35°C) and high temperature storage (35-40°C). For each type of scenario, a different pH value control strategy is set. Under room temperature storage conditions, a wider pH control range (6.0-8.0) can be adopted; under medium temperature storage conditions, the pH needs to be controlled within the range of 6.5-7.5; and under high temperature storage conditions, it is recommended to control the pH value within the range of 7.0-7.5.

[0414] The decision rules also include adjustment strategies for environmental humidity. Experimental data show that under high humidity conditions (relative humidity>80%), the pH control range needs to be further narrowed to improve stability. Specifically, under high temperature and high humidity conditions, it is recommended to strictly control the pH value within the range of 7.2±0.2 and increase the amount of protective agent used.

[0415] S6.4 Rule Verification:

[0416] Verify the accuracy of the classification decision rules based on experimental data.

[0417] In order to verify the effectiveness of the decision rule, a systematic verification experiment was designed in the embodiment of the present application. First, the verification was carried out on a laboratory scale, and typical conditions of different temperature and humidity combinations were selected to prepare formula samples that met the requirements of the decision rule for storage tests. The results showed that the formula optimized according to the decision rule had a bacterial viability retention rate of more than 90% at room temperature for 4 months, and a retention rate of more than 80% under high temperature conditions.

[0418] Further expansion to pilot scale validation, using 100kg batches to confirm the reliability of the decision rule. The validation process paid special attention to the uniformity and stability of the formula. Through multi-point sampling analysis, it was confirmed that the coefficient of variation of the uniformity of the formula throughout the batch was controlled within 5%. The results of the long-term storage test were in good agreement with the results of the laboratory scale, confirming the practicality of the decision rule.

[0419] S6.5 parameter adjustment:

[0420] Dynamically adjust probability distribution model parameters.

[0421] Dynamic adjustment of model parameters is the key to ensuring prediction accuracy. The embodiment of the present application establishes a parameter adjustment mechanism based on feedback: first, each batch of products is tracked and monitored to collect actual bacterial activity change data; then, the measured data is compared with the model prediction value to calculate the prediction deviation; finally, the model parameters are dynamically adjusted according to the size and direction of the deviation.

[0422] The parameter adjustment adopts an adaptive algorithm, and the adjustment step is proportional to the prediction deviation. For example, when the prediction deviation exceeds 10%, the temperature sensitivity coefficient β needs to be adjusted; when the deviation exceeds 15%, the distribution width parameter σ needs to be adjusted at the same time. Through this dynamic adjustment mechanism, the prediction accuracy of the model is significantly improved, and the mean prediction deviation is reduced to less than 8%.

[0423] S7: Stability Verification Program

[0424] S7.1 Normal temperature and humidity verification:

[0425] Store at normal temperature and humidity conditions (25°C, relative humidity 45%) for 4 months, and test bacterial activity once a month.

[0426] S7.2 Normal temperature and high humidity verification:

[0427] Store at room temperature and high humidity conditions (25°C, relative humidity 85%) for 4 months, and test bacterial activity once a month.

[0428] S7.3 High temperature and humidity verification:

[0429] The samples were stored under high temperature and normal humidity conditions (40°C, relative humidity 45%) for 4 months, and the bacterial activity was tested once a month.

[0430] S7.4 High temperature and high humidity verification:

[0431] The samples were stored under high temperature and high humidity conditions (40°C, relative humidity 85%) for 4 months, and the bacterial activity was tested once a month.

[0432] S7.5 Trend Analysis:

[0433] Record and analyze the changing trends of bacterial activity under different storage conditions.

[0434] like Figure 5 Storage verification under normal temperature and humidity conditions (25°C, relative humidity 45%) is a basic test item. The prepared samples are placed in a constant temperature and humidity chamber, and the environmental parameters are recorded every 30 minutes, and a complete bacterial activity test is performed once a month. The experimental results show that under such conditions, the optimized formula can maintain a high bacterial activity stability, and the bacterial activity level after 4 months can still be maintained above 3.50E+08CFU / g, far exceeding the target value of 2.00E+08CFU / g.

[0435] The verification under normal temperature and high humidity conditions (25°C, relative humidity 85%) focuses on the effect of humidity on bacterial activity. Under such conditions, samples are more likely to clump, and special attention should be paid to the moisture-proof performance of packaging materials. Through experimental observation, it was found that high humidity would accelerate the attenuation of bacterial activity, but the optimized formula can still maintain good stability, and the bacterial activity level after 4 months remained above 3.00E+08CFU / g.

[0436] The verification under high temperature and normal humidity conditions (40°C, relative humidity 45%) mainly examines the effect of temperature stress on bacterial activity. This condition is a great challenge for the survival of microorganisms. Experimental data show that under high temperature conditions, the protective agent in the optimized formula plays an important role and can significantly slow down the decay rate of bacterial activity. Although the bacterial activity level has decreased, it can still be maintained above 2.50E+08CFU / g after 4 months.

[0437] The most stringent test conditions are high temperature and high humidity environments (40°C, 85% relative humidity). Under these conditions, the synergistic effect of temperature and humidity will accelerate the inactivation of microorganisms. By adopting a special protective formula and improved packaging methods, the optimized product can still maintain good stability under such extreme conditions, and the bacterial activity level after 4 months is maintained at more than 2.20E+08CFU / g.

[0438] The trends of bacterial activity changes under all storage conditions are fully recorded and analyzed. By comparing the decay curves under different conditions, we can clearly see the influence of environmental factors on product stability. These data not only verify the effectiveness of the formula, but also provide important reference for subsequent optimization.

[0439] S8: Bacterial activity detection method

[0440] When using the standard plate count method to determine the total number of colonies, it is necessary to strictly control the operating specifications of each link. First, perform a series of dilutions on the sample and select the appropriate dilution so that the number of colonies on each plate is between 30-300. The use of an automatic colony counting system can improve the accuracy and efficiency of counting. The system is equipped with a high-resolution camera and an intelligent recognition algorithm that can accurately distinguish and count various types of colonies.

[0441] When calculating the bacterial viability retention rate, the colony count measured each time is compared with the initial bacterial viability value. This ratio reflects the stability of the product during storage. In order to ensure the reliability of the data, 3 parallel samples are required for each sample, and the average value is taken as the final result. Experiments show that the bacterial viability retention rate of the optimized formula at room temperature for 4 months can generally reach more than 80%.

[0442] The pH value is monitored using a high-precision digital pH meter, which needs to be calibrated with a standard buffer before each measurement. The change in pH not only reflects the stability of the product, but is also an important indicator for judging the effectiveness of protective agents and preservatives. Experimental data show that the optimized formula can effectively buffer pH fluctuations, and the change in pH during storage is usually controlled within ±0.5 units.

[0443] The remaining effective content of the protective agent and preservative is analyzed by high performance liquid chromatography. By establishing a standard curve, the content of these active ingredients can be accurately quantified. This analysis can reveal the attenuation law of the protective system and provide a basis for further optimization of the formula. The experimental results show that within a storage period of 4 months, the effective ingredient retention rate of the protective agent is more than 70%, and the effective ingredient retention rate of the preservative is more than 80%.

[0444] S8.1 Total number determination:

[0445] The total number of colonies was determined by standard plate count method.

[0446] S8.2 Retention rate calculation:

[0447] Calculate the retention rate relative to the initial bacterial activity.

[0448] S8.3 pH monitoring:

[0449] Determine the current pH value and calculate its change.

[0450] S8.4 Content analysis:

[0451] Analyze the remaining effective content of protective agents and preservatives.

[0452] S9: Dynamic adjustment mechanism of ratio parameters

[0453] like Figure 6When conducting deviation analysis, the focus is on the difference between the actual test results and the model prediction value. By establishing a standard difference evaluation system, the deviation is divided into three levels: acceptable range (<10%), range requiring attention (10-20%), and range requiring adjustment (>20%). When significant deviations are found, the cause needs to be analyzed in depth, which may be a sudden change in environmental factors or a stability problem of the formula itself.

[0454] The evaluation of pH stability requires comprehensive consideration of multiple aspects. The first is the changing trend of pH value. The pH change curve is drawn through continuous monitoring data to analyze its fluctuation characteristics. The second is the impact of pH change on bacterial activity, and the corresponding relationship between the pH change and the bacterial activity decrease rate is established. Experimental data show that when the pH value changes by more than 0.5 units, the bacterial activity will decrease significantly. In addition, the reversibility of pH changes must also be considered. In some cases, pH fluctuations can be repaired by the buffering effect of the system itself.

[0455] When adjusting the amount of protective agent added according to the evaluation results, a gradual adjustment strategy is adopted. For situations where the protective effect needs to be increased, the amplitude of each adjustment is controlled within 10% of the original ratio, and the effect is observed for at least one month after the adjustment. If the effect is not obvious, make the next adjustment. This gradual adjustment can avoid the negative impact caused by excessive adjustment. During the adjustment process, special attention should be paid to the ratio relationship between the components of the protective agent to ensure that their synergistic effect is not affected.

[0456] The adjustment of the amount of preservatives added needs to be more cautious, because excessive preservatives may inhibit the activity of beneficial bacteria. The adjustment strategy adopts the principle of "small steps and fast progress", with each adjustment not exceeding 5%, and a rapid bacterial activity test is required to confirm the impact of the adjustment. When the preservative effect is found to be insufficient, in addition to considering adjusting the amount of addition, it is also necessary to evaluate whether the type or combination of preservatives needs to be changed.

[0457] Parameter updating is the last step in formula optimization and is also the key to ensuring the continued effectiveness of the entire optimization system. The update content includes model parameters, threshold settings, and adjustment strategies. Specifically, the various parameters in the model should be updated according to the new experimental data, especially those that reflect the influence of environmental factors; the second is to adjust the settings of various thresholds based on practical experience to make them more in line with actual production needs; and finally, the adjustment strategy should be improved to form a more refined and effective optimization plan.

[0458] This dynamic adjustment mechanism forms a complete closed-loop system that can continuously optimize formula parameters based on real-time monitoring data. In this way, not only can the stability of the product be maintained, but it can also adapt to changes in different environmental conditions and ensure the quality of the microbial fertilizer. Practice has proved that the stability of bacterial activity of products using this dynamic adjustment mechanism is significantly better than that of products using fixed formulas.

[0459] S9.1 Deviation Analysis:

[0460] Analyze the deviation between bacterial activity test results and expected targets.

[0461] Deviation analysis uses a multi-level evaluation system. The first is basic deviation analysis, which directly compares the measured bacterial activity value with the model prediction value to calculate the relative deviation. When the deviation is less than 10%, the formula performance is considered to be within the normal range; when the deviation is between 10-20%, it is necessary to enter the observation period and increase the detection frequency; when the deviation exceeds 20%, the alarm mechanism is activated and the formula needs to be adjusted immediately. By setting up this graded response mechanism, stability problems can be discovered and resolved in a timely manner.

[0462] In-depth deviation analysis also includes trend analysis. Through multiple consecutive test results, time series analysis methods are used to identify the trend characteristics of bacterial activity changes. For example, if three consecutive tests show that bacterial activity is showing an accelerating downward trend, even if the single deviation does not exceed the threshold, preventive adjustments are required. This trend analysis can detect potential stability problems in advance and prevent drastic fluctuations in bacterial activity.

[0463] S9.2 pH stability assessment:

[0464] The effect of pH stability on bacterial activity was evaluated.

[0465] The assessment of pH stability needs to consider multiple dimensions. The first is the absolute change in pH, which is determined by continuous monitoring. Experimental data show that when the pH value changes by more than ±0.2 units, special attention is required; when the change exceeds ±0.5 units, it usually leads to a significant decrease in bacterial activity. The second is the rate of change of pH. Rapid pH changes have a greater impact on bacterial activity than slow changes.

[0466] During the evaluation process, the reversibility of pH changes also needs to be considered. In some cases, pH fluctuations are temporary and the system has the ability to self-regulate. For example, in a formulation with a complete buffer system, short-term pH fluctuations will recover on their own within 24-48 hours. This self-healing ability is an important indicator for evaluating the stability of a formulation. By establishing a pH change curve, the buffering capacity and self-healing ability of the formulation can be accurately determined.

[0467] S9.3 Protective agent adjustment:

[0468] Adjust the amount of protective agent added according to the evaluation results.

[0469] The adjustment of protective agents follows the principle of "prudent and gradual". For situations where the protective effect needs to be improved, a progressive adjustment strategy is adopted. The first adjustment range is usually controlled within 5% of the original ratio. After each adjustment, it is observed for at least 2 weeks to confirm the effect before deciding whether further adjustment is needed. If the first adjustment effect is not obvious, the second adjustment can be increased to 8%, but the total adjustment range in a single cycle shall not exceed 15%.

[0470] During the adjustment process, special attention should be paid to the ratio between the components of the protective agent. For example, when using a composite protective agent, the synergistic effect of each component needs to be maintained. Experiments have found that there is an optimal ratio range between some protective agent components, and deviation from this range will reduce the protective effect. Therefore, during the adjustment, it is necessary to change the addition amount of each component synchronously to maintain their optimal ratio.

[0471] S9.4 Preservative adjustment:

[0472] Adjust the amount of preservative added based on the evaluation results.

[0473] The adjustment strategy for preservatives is more cautious, adopting the principle of "small amount and multiple times". The adjustment range is strictly controlled within 3% each time, and a rapid bacterial activity test is required after the adjustment, including a 72-hour accelerated test and a 7-day conventional test. The effectiveness of the adjustment can only be confirmed if both tests show safety. If any adverse signs are found, immediately return to the previous stable ratio.

[0474] The adjustment of preservatives also needs to consider the impact of seasonal factors. In hot seasons, due to the enhanced activity of microorganisms, it may be necessary to increase the amount of preservatives appropriately; while in cold seasons, the amount can be appropriately reduced. This seasonal adjustment is usually carried out within the range of ±5%, and the specific adjustment amount needs to be determined in combination with historical data and real-time monitoring results.

[0475] S9.5 parameter update:

[0476] Update relevant parameters in the recipe optimization model.

[0477] The parameter update adopts a progressive learning mechanism. After each new experimental data is obtained, the data quality is first evaluated to remove obviously abnormal data points. Then the valid data is input into the model and the model parameters are updated using the sliding window method. The window size is usually set to 90 days, which ensures the sensitivity of the model to new data while maintaining sufficient historical information.

[0478] An anti-oscillation mechanism is set during the parameter update process. If the new data causes the parameter fluctuation to exceed the preset threshold (usually 10%), a secondary verification process will be initiated. The update operation is performed only when multiple consecutive sets of data support parameter adjustment. This mechanism effectively prevents drastic fluctuations in parameters caused by individual abnormal data. At the same time, the updated parameters need to be cross-validated to confirm their predictive performance to ensure the stability and reliability of the model.

[0479] like Fig. 9 As shown, after adopting the method of the embodiment of the present application, under normal temperature conditions, the general NPK+ bacteria macro-element water-soluble fertilizer can maintain a decay rate of 50% within 4 months; under high temperature conditions, the alkaline fertilizer can maintain a decay rate of 50% within 4 months; if there is a region that requires high temperature storage, it is necessary to adjust the formula to make the pH value alkaline.

[0480] During the operation of the entire system, all adjustment records are fully stored in the database to form a historical archive of product optimization. These data can not only be used for retrospective analysis, but also provide important references for the development of new products. Through statistical analysis of a large number of optimization cases, the embodiments of the present application continue to accumulate experience and improve the accuracy and efficiency of formula optimization.

[0481] The present application also provides a microbial fertilizer protective agent and preservative optimization ratio system, which is used to implement the above-mentioned microbial fertilizer protective agent and preservative optimization ratio method, and the system includes:

[0482] Edge computing nodes, including:

[0483] Digital temperature sensor with an accuracy of ±0.1°C and an operating range of 20-40°C;

[0484] Capacitive humidity sensor with an accuracy of ±1%RH and a detection range of 0-100%;

[0485] Digital pH sensor with an accuracy of ±0.01pH and a measurement range of 3-9pH;

[0486] Signal conditioning circuits for sensor data acquisition;

[0487] A microprocessor for data preprocessing;

[0488] Industrial-grade programmable controller;

[0489] Industrial Ethernet communication module;

[0490] Edge computing processor using ARM architecture;

[0491] Cloud computing servers, including:

[0492] Computing servers for data analysis;

[0493] Distributed database for data storage;

[0494] Laboratory testing unit, including:

[0495] Automatic colony counting system;

[0496] Digital microscope with image analysis capabilities;

[0497] a spectrophotometer for measuring bacterial density;

[0498] Environmental test chamber with controlled temperature and humidity;

[0499] Sterile operating table.

[0500] The present application also provides a computer device, the computer device comprising:

[0501] at least one processor; and,

[0502] a memory communicatively connected to the at least one processor; wherein,

[0503] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for optimizing the ratio of the microbial fertilizer protective agent and the preservative.

[0504] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the method for optimizing the ratio of the microbial fertilizer protective agent and the preservative.

[0505] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the method for optimizing the ratio of the above-mentioned microbial fertilizer protective agent and the preservative.

[0506] The above technical solution achieves accurate optimization of the ratio of microbial fertilizer protective agent and preservative through systematic data collection, scientific model construction and intelligent dynamic adjustment, and significantly improves the stability and applicability of the product. This solution not only solves the problems existing in the existing technology, but also provides new ideas for the technological progress of the microbial fertilizer industry.

Claims

1. A method for optimizing the ratio of a microbial fertilizer protective agent and a preservative, characterized in that: Applicable to systems including cloud computing servers, edge computing nodes, and laboratory testing units, including: Collecting microbial data through the edge computing node, the microbial data including temperature, humidity and pH parameters of the storage environment, and bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions, wherein the collection range of the temperature parameter is 20-40°C, the collection range of the humidity parameter is 0-100% relative humidity, the collection range of the pH parameter is 3-9, and the bacterial activity data includes monthly colony counts and bacterial survival rates, which are collected by the laboratory detection unit; The edge computing node uploads the collected microbial data to a cloud computing server; The cloud computing server constructs a formula optimization model based on the temperature, humidity, pH value parameters and the bacterial activity data; The cloud computing server calculates and outputs the optimal ratio scheme according to the formula optimization model, and controls the edge computing node to monitor and verify the bacterial activity stability of the optimal ratio scheme under different storage conditions, wherein the optimal ratio scheme has a protective agent concentration range of 0-5‰ and a preservative concentration range of 0-2‰; The cloud computing server receives the bacterial activity detection data regularly uploaded by the edge computing node, performs stability analysis, and dynamically adjusts the ratio parameters of the protective agent and the preservative according to the analysis results.

2. The method according to claim 1, characterized in that The microbial data is collected through the edge computing node, and the microbial data includes bacterial activity data of microbial fertilizers with multiple different NPK formulas under different environmental conditions, including: The colony counts of the two basic formulations, 20-20-20+TE and 12-6-40+TE, were performed using the standard plate count method; In each basic formula, Bacillus subtilis and Bacillus amyloliquefaciens were mixed in a ratio of 1:1, and the initial amount of bacteria added was 4.00E+08 CFU / g; Monitor colony counts and pH changes monthly; Analyze bacterial survival under different pH conditions; Correlation data between environmental condition parameters and bacterial activity were recorded.

3. The method according to claim 1, characterized in that The recipe optimization model includes a high-dimensional function approximation model, and the recipe optimization model is constructed, including: Temperature, humidity, pH value, storage time, protective agent concentration, and preservative concentration were used as input variables to construct a high-dimensional feature space; Identify effective dimensions that significantly affect bacterial activity from the input variables; Based on the effective dimensions, the model complexity is reduced to construct a simplified model; Verifying the prediction accuracy of the simplified model through experimental data; The simplified model parameters are optimized according to the verification results.

4. The method according to claim 3, characterized in that The step of identifying effective dimensions that significantly affect bacterial activity from the input variables, reducing model complexity based on the effective dimensions, and constructing a simplified model includes: Determine temperature as the first dominant dimension and establish the temperature influence function; The correlation between pH value and bacterial activity was analyzed, and pH value was determined as the second dominant dimension; Establish a progressive decay model of storage time as the third dominant dimension; The simplified model is constructed by integrating three dominant dimensions; The simplified model parameters are calculated based on experimental data.

5. The method according to claim 3, characterized in that: The step of calculating and outputting the optimal ratio scheme according to the recipe optimization model includes: Classify the bacterial activity data under different formulation conditions and generate bacterial activity data sets within different pH ranges; Based on the bacterial activity data set, a two-dimensional distribution function is generated on a pH-bacterial activity plane and data projection is performed to obtain projection data; Performing Radon transformation on the projection data to obtain a transformed distribution function; A regularization term is introduced into the transformed distribution function to deal with the uncertainty caused by the limited number of samples, and a regularized distribution estimate is obtained; Based on the regularized distribution estimation, a probability distribution model for predicting bacterial activity levels under different pH conditions is constructed; The optimal matching scheme is calculated and output according to the probability distribution model.

6. The method according to claim 5, characterized in that The constructing of the probability distribution estimation model comprises: Estimate the probability distribution of maintaining the target bacterial viability level under normal temperature conditions, wherein the normal temperature condition is 25°C; Estimating the probability distribution of maintaining the target bacterial viability level under high temperature conditions, wherein the high temperature condition is 40°C; Establish classification decision rules for recipe optimization based on probability distribution; Verify the accuracy of classification decision rules based on experimental data; Dynamically adjust probability distribution model parameters.

7. The method according to claim 1, characterized in that The stability analysis includes: verifying the bacterial activity stability of the optimal ratio solution under different storage conditions, specifically including: Store at room temperature and humidity for 4 months, and test bacterial activity once a month. The room temperature and humidity conditions are 25°C and 45% relative humidity. Store at room temperature and high humidity for 4 months, and test the bacterial activity once a month. The room temperature and high humidity conditions are 25°C and 85% relative humidity; Store under high temperature and normal humidity conditions for 4 months, and test the bacterial activity once a month. The high temperature and normal humidity conditions are 40°C and 45% relative humidity; Store under high temperature and high humidity conditions for 4 months, and test the bacterial activity once a month. The high temperature and high humidity conditions are 40°C and 85% relative humidity; Record and analyze the changing trends of bacterial activity under different storage conditions.

8. The method according to claim 7, characterized in that The monthly bacterial activity test includes: The total number of colonies was determined by the standard plate count method; Calculate the retention rate relative to the initial bacterial activity; Determine the current pH value and calculate its change; Analyze the remaining effective content of protective agents and preservatives.

9. The method according to claim 1, characterized in that: The parameters for dynamically adjusting the ratio of the protective agent to the preservative include: Analyze the deviation between bacterial activity test results and expected targets; Evaluate the effect of pH stability on bacterial activity; Adjust the amount of protective agent added according to the evaluation results; Adjust the amount of preservatives added based on the evaluation results; Update relevant parameters in the recipe optimization model.

10. A system for optimizing the ratio of microbial fertilizer protective agent and preservative, characterized in that: A method for optimizing the ratio of a microbial fertilizer protective agent to a preservative according to any one of claims 1 to 9, the system comprising: Edge computing nodes, including: Digital temperature sensor with an accuracy of ±0.1°C and an operating range of 20-40°C; Capacitive humidity sensor with an accuracy of ±1%RH and a detection range of 0-100%; Digital pH sensor with an accuracy of ±0.01pH and a measurement range of 3-9pH; Signal conditioning circuits for sensor data acquisition; A microprocessor for data preprocessing; Industrial-grade programmable controller; Industrial Ethernet communication module; Edge computing processor using ARM architecture; Cloud computing servers, including: Computing servers for data analysis; Distributed database for data storage; Laboratory testing unit, including: Automatic colony counting system; Digital microscope with image analysis capabilities; a spectrophotometer for measuring bacterial density; Environmental test chamber with controlled temperature and humidity; Sterile operating table.

Images