Polygonatum sibiricum bacterium enzyme synergistic fermentation intelligent control optimization method and system and storage medium

Abstract

The invention relates to the field of microbial fermentation engineering, and discloses an intelligent control optimization method and system for polygonatum sibiricum bacterium-enzyme synergistic fermentation and a storage medium. The method comprises the following steps: collecting fermentation parameter real-time data, and analyzing flora imbalance and enzyme system imbalance index generation system state description; preprocessing and extracting feature vectors, and analyzing correlation to determine core influence factors; calculating parameter coupling interaction strength through a support vector machine; when the intensity exceeds the standard, simulating a multi-parameter coordination scene to generate a potential adjustment scheme and enzyme system collaborative prediction data, and screening candidate schemes; evaluating a correction effect of the candidate scheme on flora imbalance by using a random forest, and determining an optimal parameter combination; adjusting parameters according to the optimal combination and calculating real-time feedback deviation; and if the deviation exceeds the limit, iteratively optimizing the pH value and the stirring rate to obtain a final scheme. According to the method, the problem that quantitative recognition and coordinated regulation of multi-parameter coupling interaction in polygonatum sibiricum bacterium enzyme synergistic fermentation are difficult is solved, and the stability of a fermentation system and the product quality are improved.

Description

Technical Field

[0001] This application relates to the field of microbial fermentation engineering, and in particular to a method, system and storage medium for intelligent control and optimization of co-fermentation of Polygonatum sibiricum enzyme. Background Technology

[0002] As an important medicinal and edible raw material, Polygonatum odoratum's active ingredients are often transformed and enhanced through the synergistic fermentation of microorganisms and enzymes, forming a Polygonatum odoratum-enzyme synergistic fermentation system. This process involves the dynamic control of multiple key parameters such as temperature, dissolved oxygen, pH, and stirring rate. These parameters have complex coupling and interaction relationships, directly affecting the balance of the microbial community structure and the synergistic activity of the enzyme system. Achieving precise and intelligent control of this system is a core technological challenge for improving the quality, yield, and process stability of Polygonatum odoratum fermentation products.

[0003] Existing methods for controlling fermentation processes have significant limitations, making it difficult to effectively address the complex dynamic problems of strong coupling of multiple parameters in the co-fermentation of Polygonatum odoratum enzymes. First, common methods often rely on empirical rules based on single variables or fixed thresholds for control, such as cooling when the temperature exceeds a preset upper limit or increasing the stirring rate when dissolved oxygen falls below a lower limit. This approach ignores the fact that temperature changes simultaneously affect dissolved oxygen solubility, microbial metabolic rate, and pH drift, while adjusting the stirring rate, in turn, affects shear force and dissolved oxygen distribution. The chain reaction between parameters can easily lead to unbalanced outcomes, causing microbial imbalance or enzyme system dysregulation. Second, some methods attempt to introduce simple feedback control, such as PID controllers for pH adjustment, but their design typically targets only a single controlled variable and cannot handle the nonlinear interactions and synergistic requirements between multiple variables such as temperature, dissolved oxygen, pH, and stirring rate. When multiple parameters simultaneously deviate from their ideal ranges, this independent control strategy cannot identify and quantify the coupling interaction strength between parameters, leading to uncoordinated or even conflicting adjustments and exacerbating system instability. Furthermore, while methods exist for prediction using data models, such as using a single regression model to predict the trend of a certain parameter, there is a lack of joint analysis and feature extraction of the two core biological state indicators, "microbial community imbalance" and "enzyme system disorder," failing to identify core influencing factors from the perspective of the overall system state. Therefore, existing technologies generally suffer from shortcomings such as delayed identification of multi-parameter coupling relationships, lack of synergy in adjustment schemes, and inability to generate coordinated contingency plans in advance when interaction intensity exceeds limits, resulting in poor stability and low efficiency in the co-fermentation process of Polygonatum odoratum enzymes.

[0004] To address the above deficiencies, this application combines real-time multi-parameter data acquisition, intelligent analysis using support vector machines and random forest models, and feedback-based iterative optimization to solve the problem of difficulty in quantifying, identifying, and coordinating the intensity of multi-parameter coupling interactions in the co-fermentation of Polygonatum odoratum enzymes. This enables early prediction and coordinated correction of microbial imbalance and enzyme system disorder, improving the stability of the fermentation system, product quality, and the overall level of intelligent control. Summary of the Invention

[0005] This application provides a method, system, and storage medium for intelligent control optimization of co-fermentation of Polygonatum sibiricum enzymes. It solves the problem of difficulty in quantifying, identifying, and coordinating the interaction intensity of multiple parameters in co-fermentation of Polygonatum sibiricum enzymes, realizes early prediction and coordinated correction of microbial imbalance and enzyme system disorder, and improves the stability of the fermentation system, product quality, and overall intelligent control level.

[0006] In a first aspect, this application provides a method for intelligent control and optimization of co-fermentation of Polygonatum sibiricum and enzymes, the method comprising: Step S101: Collect real-time data on temperature, dissolved oxygen, pH and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system, analyze the microbial imbalance index and enzyme system disorder index based on the real-time data, and generate an overall state description of the system. Step S102: Preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and determine the core influencing factors; Step S103: Input the core influencing factors into the pre-trained support vector machine model and calculate the parameter coupling interaction strength value; Step S104: If the parameter coupling interaction strength value exceeds the preset strength threshold, simulate the multi-parameter coordinated change scenario to generate potential adjustment schemes, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes. Step S105: Input the enzyme system co-maintenance prediction data into a pre-trained random forest model, evaluate the impact of the candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination; Step S106: Adjust the fermentation system operating parameters according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the predicted data of enzyme system synergistic maintenance corresponding to the optimal parameter combination. Step S107: If the deviation value exceeds the preset deviation range, the pH value and stirring rate are updated through an iterative optimization algorithm to obtain the final coordinated change scheme.

[0007] Secondly, this application provides an intelligent control and optimization system for the co-fermentation of Polygonatum sibiricum enzymes, used to implement the aforementioned intelligent control and optimization method for the co-fermentation of Polygonatum sibiricum enzymes, the system comprising: The data acquisition and analysis module collects real-time data on temperature, dissolved oxygen, pH value and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system. Based on the real-time data, it analyzes the microbial imbalance index and enzyme system disorder index and generates an overall description of the system's state. The influencing factor determination module is used to preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and determine the core influencing factors. The coupling interaction calculation module is used to input the core influencing factors into a pre-trained support vector machine model and calculate the parameter coupling interaction strength value. The candidate scheme screening module is used to simulate a multi-parameter coordinated change scenario to generate potential adjustment schemes when the parameter coupling interaction strength value exceeds a preset strength threshold, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes. The optimal parameter determination module is used to input the enzyme system co-maintenance prediction data into a pre-trained random forest model, evaluate the impact of the candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination. The feedback deviation calculation module is used to adjust the operating parameters of the fermentation system according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination. The parameter optimization and adjustment module is used to update the pH value and stirring rate through an iterative optimization algorithm when the deviation value exceeds the preset deviation range, so as to obtain the final coordinated change scheme.

[0008] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is run by a processor, the processor executes the intelligent control and optimization method for co-fermentation of Polygonatum sibiricum enzyme.

[0009] This application proposes a method, system, and storage medium for intelligent control optimization of Polygonatum sibiricum enzyme co-fermentation. It solves the problem of difficulty in quantifying, identifying, and coordinating the interaction strength of multiple parameters in Polygonatum sibiricum enzyme co-fermentation, achieving early prediction and coordinated correction of microbial imbalance and enzyme system disorder, and improving the stability of the fermentation system, product quality, and overall intelligent control level. Compared with existing technologies, the beneficial effects of this application's technical solution are at least as follows: First, by collecting real-time data on temperature, dissolved oxygen, pH, and stirring rate, and calculating microbial imbalance and enzyme system disorder indicators based on dynamic datasets, a comprehensive and quantitative characterization of the fermentation system's operating status can be generated, providing an accurate data foundation for subsequent intelligent analysis.

[0010] Second, by extracting feature vectors and analyzing the correlation between microbial imbalance and enzyme system disorder, the core influencing factors can be identified, and the key parameters that lead to system instability and their interactions can be accurately identified, overcoming the shortcomings of existing technologies that isolate the analysis of a single parameter and ignore the coupling effect of multiple variables.

[0011] Third, by inputting the core influencing factors into the support vector machine model to calculate the coupling interaction strength value of the parameters, an early warning can be given when the interaction strength exceeds the threshold, triggering the simulation of multi-parameter coordinated change scenarios and the generation of potential adjustment schemes, thus realizing the proactive identification and contingency plan preparation for strong coupling risks.

[0012] Fourth, by analyzing the impact of potential adjustment schemes on enzyme synergy and generating prediction data for enzyme synergy maintenance, and by using a random forest model to evaluate the impact of candidate schemes on community imbalance correction, the optimal parameter combination that balances community balance and enzyme synergy can be screened, thus achieving multi-objective collaborative optimization.

[0013] Fifth, by adjusting parameters according to the optimal parameter combination and collecting real-time feedback data to calculate the deviation value, and updating the pH value and stirring rate through iterative optimization algorithm when the deviation exceeds the limit, a final coordinated change scheme is formed, and a closed-loop control mechanism of "prediction-adjustment-feedback-fine-tuning" is constructed, which improves the accuracy of regulation and the long-term stability of the system. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the intelligent control optimization method for co-fermentation of Polygonatum sibiricum and enzymes in this application. Figure 2 This is a graph showing the relationship between the out-of-bag error of the random forest and the number of decision trees in this application; Figure 3 This is a plot of the particle swarm iteration curve in this application; Figure 4 This is a comparison chart of the overall performance in this application; Figure 5 This is a comparison chart of parameter adjustment accuracy in this application; Figure 6 This is a comparison chart of the technical advantages in this application; Figure 7 This is a schematic diagram of the intelligent control and optimization system for co-fermentation of Polygonatum sibiricum enzymes in this application. Detailed Implementation

[0016] This application provides a method, system, and storage medium for intelligent control optimization of co-fermentation with Polygonatum sibiricum enzymes. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent control and optimization method for co-fermentation of Polygonatum odoratum and enzymes in this application includes: Step S101: Collect real-time data on temperature, dissolved oxygen, pH and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system. Analyze the microbial imbalance index and enzyme system disorder index based on the real-time data to generate an overall description of the system's state.

[0018] In one specific embodiment, step S101 includes the following steps: Real-time data on temperature, dissolved oxygen, pH, and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were collected and integrated to form a dynamic dataset. The temperature change rate and dissolved oxygen fluctuation amplitude are calculated based on dynamic datasets, and the ratio of the temperature change rate to the dissolved oxygen fluctuation amplitude is used as an indicator of microbial community imbalance. pH deviation and stirring rate adjustment frequency are calculated based on dynamic datasets, and the product of pH deviation and stirring rate adjustment frequency is used as an indicator of enzyme system imbalance. Based on microbial community imbalance indicators and enzyme system disorder indicators, a description of the overall state of the fermentation system is generated to characterize its operation.

[0019] Specifically, during the co-fermentation of Polygonatum odoratum and its enzymes, temperature, dissolved oxygen, pH, and stirring rate data were collected every minute using temperature sensors, dissolved oxygen sensors, pH meters, and a stirring rate inverter. Simultaneously, parameters for the pretreatment of Polygonatum odoratum raw materials and the parameters for the enzyme system were also collected. The raw material pretreatment parameters included drying at 50℃, pulverizing through a 40-mesh sieve, and sterilization at 0.1 MPa / 121℃ for 20 minutes. The enzyme system parameters included the types and volume fractions of fermenting bacteria (2%–4%), and the types and amounts of enzymes added (1%–3% of the raw material mass). These data were indexed and integrated according to timestamps to form a dynamic dataset with the collection time as the row dimension and various parameters as the column dimension. This dataset contains the dynamic change trajectory of each parameter during the fermentation process, providing raw data support for subsequent indicator calculations.

[0020] When calculating microbial community imbalance indicators based on dynamic datasets, temperature data from two consecutive adjacent time points are first selected. Let the temperature at the previous time point be T1, the temperature at the current time point be T2, and the time interval be... (Fixed to 1 minute), using the formula The temperature change rate (unit: °C / min) is calculated; simultaneously, within a set statistical period (e.g., 10 minutes), the maximum dissolved oxygen data within that period is extracted. and minimum value Through formula The dissolved oxygen fluctuation range (unit: mg / L) is calculated. Then, the temperature change rate and dissolved oxygen fluctuation range are substituted into the ratio formula R1 = temperature change rate / dissolved oxygen fluctuation range to obtain the microbial community imbalance index R1. This index quantifies the correlation between dynamic temperature change and dissolved oxygen stability, directly reflecting the sensitivity of the microbial community to environmental changes and the balance state of the microbial community structure. For example, when the temperature change rate is 0.3℃ / min and the dissolved oxygen fluctuation range is 1.5mg / L, R1 = 0.2, indicating that the microbial community is in a slightly unbalanced state.

[0021] When calculating enzyme system imbalance indicators, first set a suitable target pH value for the fermentation system. (Set to pH 6.5 based on the co-fermentation characteristics of Polygonatum sibiricum enzymes), select the pH value at the current sampling time. Through formula The pH deviation (unitless) was calculated. Simultaneously, within the same 10-minute statistical period, the number of times the stirring rate was adjusted (N) was recorded. The stirring rate adjustment frequency (times / minute) was calculated using the formula N / 10. Then, the pH deviation and stirring rate adjustment frequency were substituted into the product formula R2 = pH deviation × stirring rate adjustment frequency to obtain the enzyme system imbalance index R2. This index quantifies the correlation between the degree of pH deviation and the stirring adjustment frequency, reflecting the degree to which enzyme activity is affected by environmental parameters. For example, when the pH deviation is 0.3 and the stirring rate adjustment frequency is 0.2 times / minute, R2 = 0.06, indicating that the enzyme system is in a state of mild imbalance.

[0022] When generating a description of the overall state of the system, the microbial imbalance index R1 and the enzyme system imbalance index R2 are matched with preset grading standards. The preset values ​​are: R1 < 0.2 and R2 < 0.05 for normal state; 0.2 ≤ R1 < 0.5 and 0.05 ≤ R2 < 0.1 for slight imbalance / imbalance; 0.5 ≤ R1 < 0.8 and 0.1 ≤ R2 < 0.2 for moderate imbalance / imbalance; and R1 ≥ 0.8 and R2 ≥ 0.2 for severe imbalance / imbalance. Combining the specific values ​​of the two indicators and their corresponding grades, a description of the overall state of the system is formed, including the indicator values, grading results, and dynamic changes in parameters. For example, "Microbial imbalance index 0.3, enzyme system imbalance index 0.08, the fermentation system is in a state of slight imbalance and mild imbalance, the temperature is slowly rising, dissolved oxygen fluctuates moderately, the pH value is slightly lower than the target value, and the stirring rate is adjusted at a low frequency."

[0023] By accurately collecting key environmental parameters and constructing a dynamic dataset, a quantitative correlation between environmental parameters and the state of microbial communities and enzyme systems was established. The microbial community imbalance index directly links the dynamic changes of the two key environmental parameters to the microbial community balance state through the ratio of temperature change rate to dissolved oxygen fluctuation amplitude. The enzyme system imbalance index quantifies the comprehensive impact of pH stability and stirring intensity on enzyme activity through the product of pH deviation and stirring rate adjustment frequency. The generated overall system state description achieves a comprehensive characterization of the fermentation system's operating status, providing a precise data foundation for subsequent identification of core influencing factors and parameter control. This solves the shortcomings of existing technologies in terms of vague judgment of fermentation system state and lack of targeted control, enabling subsequent control to be carried out based on clear biological state indicators, avoiding the aggravation of microbial community imbalance or enzyme system imbalance caused by blind adjustment.

[0024] Step S102: Preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and identify the core influencing factors.

[0025] In one specific embodiment, step S102 includes the following steps: Preprocess the data in the overall system state description to obtain standardized state data; Based on standardized state data, principal component analysis was used to extract feature vectors of microbial community imbalance, enzyme system disorder, and environmental parameters to construct a feature dataset. For the feature dataset, the correlation coefficient between the microbial imbalance feature vector and the enzyme system imbalance feature vector is calculated, and high coefficient indicators with an absolute value of correlation coefficient greater than a preset correlation threshold are selected. A decision tree model is constructed based on high-coefficient indicators. The decision tree model is then pruned to identify the core influencing factors.

[0026] Specifically, when preprocessing the data in the overall system state description, outliers need to be screened and removed first. The 3σ criterion is used to identify outliers deviating from the data distribution; that is, the mean μ and standard deviation σ of each data sequence are calculated, and data exceeding the range [μ-3σ, μ+3σ] are identified as outliers and removed. This processing procedure is performed separately for microbial imbalance indicators and enzyme system imbalance indicators corresponding to temperature, dissolved oxygen, pH, and stirring rate. Subsequently, the data after outlier removal undergoes a standardization transformation using the Z-score standardization method, calculated using the formula... Calculate standardized state data, where For standardized data, The original data, The mean of the original data sequence. The standard deviation of the original data sequence is 0. After this processing, the mean of all standardized state data is 0 and the standard deviation is 1, eliminating the influence of differences in the dimensions and units of different parameters. For example, the original value of the temperature-related microbial imbalance index is 0.3, and its mean is 1. 0.25, standard deviation The value is 0.1, and the standardized data is (0.3-0.25) / 0.1=0.5.

[0027] Based on standardized state data, principal component analysis (PCA) was used to extract eigenvectors. First, a standardized data matrix was constructed, with rows representing the number of data collections and columns containing data related to microbial imbalance, enzyme system disorder, and four environmental parameters: temperature, dissolved oxygen, pH, and stirring rate. The covariance matrix was calculated, and then eigenvalue decomposition was performed to solve the characteristic equation, yielding eigenvalues ​​and their corresponding eigenvectors. The eigenvalues ​​were sorted from largest to smallest, and the eigenvectors corresponding to the top m eigenvalues ​​with a cumulative contribution rate of 85% were selected as principal components. Based on the original data weight distribution corresponding to the principal components, the principal components with a weight exceeding 60% related to microbial imbalance were defined as microbial imbalance eigenvectors; those with a weight exceeding 60% related to enzyme system disorder were defined as enzyme system disorder eigenvectors; and those with a weight exceeding 60% related to environmental parameters were defined as environmental parameter eigenvectors. These three components together constitute the feature dataset, achieving data dimensionality reduction while preserving core information.

[0028] For the feature dataset, the correlation coefficient between the microbial imbalance feature vector and the enzyme system imbalance feature vector is calculated using the Pearson correlation coefficient method. The correlation coefficient ranges from -1 to 1, with the absolute value closer to 1 indicating a stronger correlation. A preset correlation threshold of 0.7 is used to select indicators with an absolute correlation coefficient greater than 0.7. These high-coefficient indicators reflect a strong association between microbial imbalance and enzyme system imbalance. For example, if a data set yields a correlation coefficient of 0.82, the corresponding indicator is included in the high-coefficient indicator set.

[0029] A decision tree model is constructed based on high-coefficient indicators. These indicators serve as input features, and the stability level of the fermentation system (divided into stable, relatively stable, and unstable levels according to the combined state of microbial imbalance and enzyme system dysregulation) is used as the output label. The training dataset consists of 1000 sets of high-coefficient indicators and their corresponding stability levels recorded during historical fermentation processes. The model is constructed using the ID3 algorithm, with information gain maximization as the feature selection criterion. The information gain formula is:

[0030] in, Let be the information entropy of dataset D, and 'a' be the feature. Let a be the set of values ​​for feature a. Take feature a in dataset D v The decision tree is generated by recursively partitioning the sample set. Partitioning stops when all samples at a node belong to the same class or the information gain is less than 0.01. The decision tree model is then pruned using a cost-complexity pruning method, incorporating pruning parameters. Calculate the cost complexity for each non-leaf node. The cost complexity formula is: ,in, It is the core evaluation index for the decision tree model of Solomon's seal enzyme co-fermentation when pruning for cost complexity; its full name is "subtree". of "Cost complexity" is used to measure the complexity of nodes. The overall merits of the rooted subtree in terms of "fitting accuracy" and "structural complexity"; For nodes The training error, For nodes The number of leaf nodes in the corresponding subtree is increased gradually. The branch with the smallest increase in cost complexity is deleted until the model's accuracy on the validation set (selecting 200 sets of historical data) no longer improves. The input features corresponding to the final retained decision tree nodes are the core influencing factors.

[0031] Preprocessing eliminates data interference and unifies dimensions, enabling subsequent analysis to be based on standardized data and avoiding analytical biases caused by differences in dimensions. Principal component analysis reduces data dimensionality, extracts key feature vectors, and reduces the impact of redundant information on the analysis. Correlation coefficient calculation screens strongly correlated indicators, focusing on the core association between microbial imbalance and enzyme system dysregulation. Decision tree model construction and pruning accurately locate core influencing factors, solving the shortcomings of vague regulation direction and lack of specificity in existing technologies. This allows subsequent parameter regulation to focus on key factors, avoids blind adjustments, and provides a clear basis for multi-parameter synergistic regulation.

[0032] Step S103: Input the core influencing factors into the pre-trained support vector machine model and calculate the parameter coupling interaction strength value.

[0033] In one specific embodiment, step S103 includes the following steps: The core influencing factors are converted into feature factor vectors and then input into a pre-trained support vector machine model. The dominant bacteria's inhibition requirement and the weak bacteria's activation potential in the Polygonatum fermentation system are classified using a support vector machine model, and the classification probability vector is output. Based on the classification probability vector, the parameter interaction coefficients between temperature, dissolved oxygen, pH value and stirring rate corresponding to each core influencing factor are calculated. The parameter coupling interaction strength value is obtained by weighting and summing the weight proportions of the classification probability vectors and the parameter interaction coefficients.

[0034] Specifically, when converting core influencing factors into feature factor vectors, it is necessary to first clarify the specific dimensions of the core influencing factors, including temperature-related influencing factors, dissolved oxygen-related factors, pH-related factors, and stirring rate-related factors. The data corresponding to each dimension are derived from the high-coefficient indicators determined in step S102, such as the temperature-dissolved oxygen coupling factor and the pH-enzyme activity correlation factor. These core influencing factors are then vectorized, with each factor corresponding to one dimension of the feature vector. The vector elements are taken as the standardized data of that factor (values ​​after Z-score standardization). A feature factor vector of dimension m×1 is constructed (m is the number of core influencing factors, determined to be 6 based on actual fermentation data). For example, the feature factor vector corresponding to a certain group of core influencing factors is... This ensures that the vector data is correlated with the parameters of the co-fermentation scenario of Polygonatum odoratum enzyme.

[0035] The support vector machine (SVM) model training requires historical data from the co-fermentation of *Polygonatum sibiricum* enzymes. The training dataset contains 5000 samples, each containing a feature vector and its corresponding label (either a label indicating the inhibitory demand of dominant bacteria or a label indicating the activation potential of weaker bacteria). The labels are determined manually by annotating the microbial community regulation needs during fermentation. For example, when dominant bacteria proliferate excessively, it is labeled as an inhibitory demand (label value 1), and when weaker bacteria have insufficient activity, it is labeled as an activation potential (label value 0). The model uses a C-SVC variant, and the kernel function is a radial basis function. ,in The parameters are set to 0.1, the penalty coefficient C is set to 1.0, and the loss function is minimized using the gradient descent algorithm. The number of iterations is set to 1000, and the learning rate is 0.001. Training stops when the model achieves a classification accuracy of over 92% on the validation set (1000 sets of historical data). After training, the feature factor vector is input into the model. The model maps the input vector to a high-dimensional feature space using radial basis functions, calculates the distance between the sample and the hyperplane, and outputs a classification probability vector. The vector contains two elements, corresponding to the probability of dominant bacteria inhibiting demand and the probability of weak bacteria activating potential, respectively. For example, an output classification probability vector of [0.75, 0.25] indicates that under this set of core influencing factors, the probability of dominant bacteria inhibiting demand is 75%, and the probability of weak bacteria activating potential is 25%.

[0036] When calculating the parameter interaction coefficients based on classification probability vectors, the four parameters corresponding to each core influencing factor—temperature, dissolved oxygen, pH, and stirring rate—are first identified. The interaction relationship between each pair of parameters is then quantified. Taking temperature (Te) and dissolved oxygen (Oe) as an example, 100 sets of parameter variation data corresponding to this group of core influencing factors from historical fermentation data are selected, and their covariance is calculated. ,in The average temperature. The mean dissolved oxygen level is used as the basis for calculating the standard deviation of temperature. and the standard deviation of dissolved oxygen Through formula The interaction coefficient between temperature and dissolved oxygen is obtained, with a value ranging from -1 to 1. A larger absolute value indicates a stronger interaction. Similarly, the interaction coefficients between temperature and pH (pHe), temperature and stirring rate (Ze), dissolved oxygen and pH, dissolved oxygen and stirring rate, and pH and stirring rate are calculated separately, forming six interaction coefficients. For example, the calculated... , , , , , .

[0037] When calculating the parameter coupling interaction strength value based on the weight proportions of the classification probability vector and the parameter interaction coefficient, first determine the weight proportions of the classification probability vector, and set the weight of the probability of the dominant bacteria inhibiting the demand as... The weight of the activation potential probability of weak bacteria is set as ,satisfy The weight values ​​are determined based on the degree of influence of the two demands on the stability of the fermentation system in historical data. First, calculate the overall weight of the microbial community's demand, using the following formula: Substituting the aforementioned classification probability vector data, we obtain Here, P1 and P2 are two elements in the classification probability vector output by the support vector machine model. Then, each parameter interaction coefficient is multiplied by this comprehensive weight to obtain the weighted interaction coefficient of a single parameter pair, for example... , , , , , Finally, the weighted parameter interaction coefficients are summed to obtain the parameter coupling interaction strength value. .

[0038] By converting core influencing factors into feature vectors, the standardization and relevance of model input data are ensured. The training of the support vector machine model is combined with historical data on the co-fermentation of Polygonatum odoratum enzymes to clarify the model parameters and scenario adaptability, achieving accurate classification of microbial community regulation needs. The calculation of parameter interaction coefficients is based on covariance and standard deviation to quantify the inherent interaction relationship between parameters. The weighted summation process integrates classification probability weights and parameter interaction coefficients to comprehensively reflect the overall strength of parameter coupling interaction, solving the defects of lagging parameter interaction identification and lack of predictive regulation in existing technologies. This provides a quantitative basis for the generation of subsequent multi-parameter coordinated adjustment schemes, ensuring timely triggering of early warning and regulation when the strength of parameter coupling interaction exceeds the standard.

[0039] Step S104: If the parameter coupling interaction strength value exceeds the preset strength threshold, simulate the multi-parameter coordinated change scenario to generate potential adjustment schemes, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes.

[0040] In one specific embodiment, step S104 includes the following steps: If the parameter coupling interaction strength value exceeds the preset strength threshold, the temperature, dissolved oxygen, pH value and stirring rate in the fermentation system are used as the control objects to simulate multiple sets of multi-parameter coordinated change scenarios and generate multiple potential adjustment schemes. The effects of each potential adjustment scheme on enzyme synergy in the Polygonatum sibiricum enzyme co-fermentation system were analyzed, and the degree of enzyme activity balance and the potential for correcting enzyme imbalance were quantified. Based on the impact analysis results, predictive data for enzyme system synergistic maintenance corresponding to each potential adjustment scheme are generated. Potential adjustment schemes that meet the preset enzyme system synergy criteria in the enzyme system synergy maintenance prediction data are selected as candidate schemes.

[0041] Specifically, the preset intensity threshold is determined based on historical stable operation data of Polygonatum odoratum enzyme co-fermentation, and is set to 1.8. When the calculated parameter coupling interaction intensity value exceeds this threshold, it indicates that the coupling interaction between temperature, dissolved oxygen, pH value and stirring rate in the system has exceeded the stable range, and multi-parameter coordination adjustment needs to be initiated. When simulating multiple coordinated change scenarios using these four parameters as the control objects, it is necessary to first clarify the control range of each parameter. The temperature control range is 28~35℃, dissolved oxygen is 4~10mg / L, pH is 5.5~7.5, and stirring speed is 150~300r / min. Each parameter is divided into 5 control nodes at equal intervals. Multiple sets of multi-parameter coordinated change scenarios are generated through full permutation and combination. For example, the combination of temperature 30℃ with dissolved oxygen 6mg / L, pH 6.5, and stirring speed 200r / min, and the combination of temperature 32℃ with dissolved oxygen 8mg / L, pH 6.8, and stirring speed 220r / min, etc., generate a total of 5×5×5×5=625 potential adjustment schemes to ensure coverage of the main possible range of parameter control.

[0042] Analyzing the impact of each potential adjustment scheme on enzyme synergy requires relying on a pre-trained enzyme synergy prediction model. This model is a deep learning-based neural network model. The input layer consists of parameter combinations of potential adjustment schemes (temperature, dissolved oxygen, pH, and stirring rate). The hidden layer contains three fully connected layers: the first layer has 64 neurons, the second has 32, and the third has 16. The output layer consists of two neurons, corresponding to the quantitative values ​​of enzyme activity balance and enzyme system imbalance correction potential, respectively. The model training dataset consists of 8000 historical data sets. Each data set includes parameter combinations and corresponding measured enzyme activity balance and enzyme system imbalance correction potential. Enzyme activity balance is calculated by detecting the activity ratio of key enzymes (such as cellulase and amylase) in the fermentation system. Enzyme system imbalance correction potential is determined by the change in enzyme system imbalance indicators before and after adjustment. The training process uses the Adam optimizer with a learning rate of 0.002, 2000 iterations, and mean squared error as the loss function. Training stops when the model's prediction error on the validation set (1600 data sets) is less than 5%. The parameter combination of each potential adjustment scheme is input into the model. The model is processed by the linear transformation of the hidden layer and the activation function (ReLU function), and outputs the enzyme activity balance degree (value range 0~1, the closer to 1, the better the balance effect) and the enzyme system imbalance correction potential (value range 0~1, the closer to 1, the stronger the correction ability). For example, the output corresponding to a certain potential adjustment scheme is enzyme activity balance degree 0.85 and enzyme system imbalance correction potential 0.82.

[0043] When generating prediction data for enzyme system synergy maintenance corresponding to each potential adjustment scheme, the enzyme activity balance and enzyme system imbalance correction potential output by the model are integrated into a two-dimensional vector, with the vector form being [enzyme activity balance, enzyme system imbalance correction potential]. For example, the prediction data for enzyme system synergy maintenance corresponding to the aforementioned scheme is [0.85, 0.82]. This data comprehensively reflects the maintenance and correction effects of the scheme on enzyme system synergy. When screening candidate schemes, the preset enzyme system synergy standard is that the enzyme activity balance is ≥0.7 and the enzyme system imbalance correction potential is ≥0.7. The prediction data for enzyme system synergy maintenance of each potential adjustment scheme are compared with this standard, and schemes that meet both conditions are retained as candidate schemes. For example, the prediction data of a potential adjustment scheme is [0.72, 0.71], which meets the preset standard and is included in the candidate scheme; the prediction data of another scheme is [0.68, 0.80], which is rejected because the enzyme activity balance does not meet the standard. The final number of candidate solutions is determined based on the actual effect of parameter combinations. For example, 45 candidate solutions are selected from 625 potential solutions to ensure that there are enough high-quality solutions to choose from in subsequent optimization.

[0044] Multiple potential adjustment schemes are generated through full permutation and combination, covering the main scenarios of parameter regulation and avoiding the chain reaction caused by adjusting a single parameter. The training of the enzyme synergy effect prediction model is combined with historical data of Polygonatum sibiricum enzyme synergistic fermentation to accurately quantify the impact of the scheme on enzyme synergy, solving the defect of fuzzy scheme effect evaluation. Pre-set clear enzyme synergy criteria are used for screening to ensure that candidate schemes have good enzyme synergy maintenance and correction capabilities, providing a high-quality foundation for determining the optimal parameter combination in the future, and overcoming the shortcomings of existing technologies such as poor targeting of adjustment schemes and easy exacerbation of system instability.

[0045] Step S105: Input the enzyme system synergistic maintenance prediction data into the pre-trained random forest model, evaluate the impact of candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination.

[0046] In one specific embodiment, step S105 includes the following steps: The prediction data of enzyme system synergistic maintenance corresponding to the candidate schemes are correlated and integrated with the microbial community imbalance indicators to form the model input dataset; Input the model input dataset into a pre-trained random forest model and output the impact coefficient of microbial imbalance correction for each candidate scheme; Based on the influence coefficient of microbial community imbalance and combined with the prediction data of enzyme system synergistic maintenance corresponding to the candidate schemes, the comprehensive optimization value of each candidate scheme is quantified. The optimal parameter combination was determined by selecting the candidate scheme with the highest comprehensive optimization value corresponding to the combination of temperature, dissolved oxygen, pH value, and stirring rate.

[0047] Specifically, when integrating the enzyme system synergistic maintenance prediction data corresponding to the candidate schemes with the microbial imbalance index, the enzyme system synergistic maintenance prediction data is the two-dimensional vector [enzyme activity balance degree, enzyme system imbalance correction potential] output in step S104, and the microbial imbalance index is the value calculated in step S101 (e.g., 0.3). Each candidate scheme corresponds to a set of integrated data. The integration method is to add the microbial imbalance index as the third dimension data to the enzyme system synergistic maintenance prediction data vector to form a three-dimensional input vector [enzyme activity balance degree, enzyme system imbalance correction potential, microbial imbalance index]. For example, if the enzyme system synergistic maintenance prediction data of a certain candidate scheme is [0.85, 0.82], and the corresponding microbial imbalance index is 0.3, then the integrated input vector is [0.85, 0.82, 0.3]. The integrated data of all candidate schemes together constitute the model input dataset. The row dimension of the dataset is the number of candidate schemes (e.g., 45 groups), and the column dimension is 3.

[0048] The random forest model training requires historical correlation data from the co-fermentation of Polygonatum sibiricum enzymes. The training dataset contains 6000 samples, each containing the aforementioned three-dimensional input vector and its corresponding label (microbial imbalance correction effect value). The label is determined by the change in the microbial imbalance index after the application of the candidate scheme in actual tests. The calculation method is the percentage difference between the corrected microbial imbalance index and the original microbial imbalance index, i.e., label value = (original index - corrected index) / original index, with a value range of 0 to 1, where a value closer to 1 indicates a better correction effect. During model construction, the number of decision trees is set to 100, with a maximum depth of 8, a minimum number of splits of 4, and a minimum number of leaf nodes of 2. The bootstrap sampling method is used to extract samples from the training dataset to construct each decision tree, with a sampling ratio of 70%. During training, each decision tree independently splits the input features using Gini impurity as the splitting criterion. The decision tree structure is generated by recursively partitioning the sample set. Splitting stops when the number of node samples is less than the minimum number of splits or when the maximum depth is reached. After all decision trees are trained, the outputs of each decision tree are aggregated through a voting mechanism to obtain the final prediction value of the model. Model validation uses 5-fold cross-validation to ensure a prediction accuracy of over 90% on the validation set, completing the training. The model input dataset is then fed into the trained random forest model, and the model outputs the microbial imbalance correction influence coefficient for each candidate solution. This coefficient ranges from 0 to 1 and directly reflects the solution's ability to correct microbial imbalance. For example, a candidate solution with an output coefficient of 0.83 indicates that the solution can correct 83% of the microbial imbalance.

[0049] When quantifying the comprehensive optimization value based on the influence coefficient of microbial imbalance correction and the prediction data of enzyme system synergy maintenance, the weight of the influence coefficient of microbial imbalance correction is set to 0.5, the weight of enzyme activity balance is set to 0.3, and the weight of enzyme system imbalance correction potential is set to 0.2, satisfying the requirement that the sum of the weights is 1. This is achieved through the formula... Calculate the overall optimization value, where A is the influence coefficient for correcting microbial imbalance, B is the degree of enzyme activity balance, and C is the potential for correcting enzyme system imbalance. For example, if a candidate scheme has A=0.83, B=0.85, and C=0.82, then... The higher the value, the better the overall optimization effect of the scheme. When selecting the optimal parameter combination, the overall optimization value of all candidate schemes is ranked, and the parameter combination of temperature, dissolved oxygen, pH value, and stirring rate corresponding to the scheme with the highest value is selected as the optimal parameter combination. For example, among 45 candidate schemes, the parameter combination corresponding to the scheme with the highest overall optimization value is temperature 32℃, dissolved oxygen 7mg / L, pH value 6.6, and stirring rate 210r / min; this combination is the finally determined optimal parameter combination.

[0050] By linking and integrating enzyme system synergy maintenance prediction data with microbial imbalance indicators, the model input data simultaneously covers enzyme system and microbial community status information, ensuring the comprehensiveness of the assessment. The training of the random forest model, combined with historical correlation data, and the aggregation prediction through multiple decision trees, improves the calculation accuracy of the microbial imbalance correction influence coefficient and solves the defect of large prediction bias of a single model. The quantification of comprehensive optimization value achieves a multi-objective balance between microbial community correction and enzyme system synergy through reasonable weight allocation, avoiding system imbalance caused by unilaterally pursuing the optimization of a single indicator. The optimal parameter combination selected in the end can simultaneously meet the needs of microbial community balance and enzyme system synergy, providing precise guidance for the stable operation of the fermentation system.

[0051] Step S106: Adjust the fermentation system operating parameters according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the predicted data of enzyme system synergistic maintenance corresponding to the optimal parameter combination.

[0052] In one specific embodiment, step S106 includes the following steps: Based on the optimal parameter combination, the temperature, dissolved oxygen, pH value and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were adjusted. Real-time feedback data on temperature, dissolved oxygen, pH, and stirring rate in the fermentation system after parameter adjustment are continuously collected and integrated into a feedback dataset according to the time series. The feedback dataset is matched with the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination according to the corresponding parameter dimensions; Based on the matching results, the deviation between the real-time feedback data and the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination is calculated.

[0053] Specifically, when adjusting the operating parameters of the fermentation system based on the optimal parameter combination, the automatic control system of the fermentation equipment executes the parameter adjustment instructions. The optimal parameter combination includes specific values ​​for temperature, dissolved oxygen, pH value, and stirring rate, such as a temperature of 32℃, dissolved oxygen of 7 mg / L, pH value of 6.6, and stirring rate of 210 r / min. After receiving the parameter instructions, the automatic control system adjusts the flow rate of the heating or cooling medium in the fermenter jacket through the temperature control module to stabilize the system temperature to the target value within 30 minutes, with fluctuations controlled within ±0.2℃; it adjusts the air intake through the aeration device, combined with real-time feedback from the dissolved oxygen sensor, to stabilize the dissolved oxygen value within the target value range of ±0.3 mg / L; it automatically adds acid or alkali solution through the acid-base titration device to adjust the pH value to the target value, controlling the deviation to not exceed ±0.1; and it adjusts the stirring speed of the impeller through the variable frequency motor to ensure that the stirring rate precisely matches the target value, with an error not exceeding ±5 r / min, ensuring the accuracy and stability of parameter adjustment.

[0054] When continuously collecting real-time feedback data after parameter adjustments, high-precision sensing equipment was used to collect temperature, dissolved oxygen, pH value, and stirring rate data once per minute for 2 hours after parameter adjustment stabilization, resulting in 120 sets of real-time data. These data were integrated in chronological order to form a time-series feedback dataset. The dataset is indexed by timestamps, with each timestamp corresponding to a set of four-dimensional parameter data. For example, timestamp 10:01 corresponds to data of [32.1℃, 6.9mg / L, 6.58, 208r / min], and timestamp 10:02 corresponds to data of [31.9℃, 7.1mg / L, 6.62, 212r / min], ensuring the temporal integrity and parameter correspondence of the data. When matching the feedback dataset with the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination by parameter dimension, the parameter dimensions included in the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination are first clarified, namely, the predicted values ​​of temperature, dissolved oxygen, pH, and stirring rate. This prediction data is the parameter-related part of the enzyme system synergistic maintenance prediction data generated in step S104 with the potential adjustment scheme corresponding to the optimal parameter combination, for example, [32℃, 7mg / L, 6.6, 210r / min]. Taking each timestamp in the feedback dataset as a unit, the real-time feedback temperature, dissolved oxygen, pH, and stirring rate are matched with the corresponding predicted values ​​by dimension, forming four sets of one-to-one corresponding parameter pairs. For example, the real-time temperature of 32.1℃ at a certain timestamp matches the predicted temperature of 32℃, and the real-time dissolved oxygen of 6.9mg / L matches the predicted dissolved oxygen of 7mg / L, ensuring that the real-time data and predicted data of each parameter dimension correspond accurately.

[0055] When calculating the deviation value based on the matching results, the root mean square error algorithm is used, and the formula is as follows: ,in Real-time feedback data for a specific parameter dimension. This is the predicted data for this parameter dimension, where n is the number of data sets (120 sets). After calculating the deviation values ​​for each of the four parameter dimensions, a weighted sum is used to obtain the overall deviation value. The weights for temperature, dissolved oxygen, pH, and stirring rate are set to 0.25, 0.3, 0.25, and 0.2, respectively, ensuring the sum of the weights equals 1. For example, if the root mean square error for temperature is 0.12℃, dissolved oxygen is 0.15mg / L, pH is 0.08, and stirring rate is 3r / min, then the overall deviation value is... .

[0056] The automatic control system precisely executes adjustment commands for the optimal parameter combination, ensuring the accuracy of parameter adjustments. Continuous real-time data collection and time-series integration fully record the dynamic changes of the adjusted system, providing a comprehensive data foundation for deviation calculation. Precise matching by parameter dimension and the application of the root mean square error algorithm enable quantitative calculation of deviation values, solving the problem of fuzzy evaluation of adjustment effects. The weighted calculation of comprehensive deviation values ​​takes into account the influence weight of different parameters on the fermentation system, making deviation assessment more in line with actual scenarios. This provides a clear quantitative basis for whether subsequent fine-tuning is needed, avoiding system fluctuations caused by blind adjustments and further ensuring the stability of the fermentation system.

[0057] Step S107: If the deviation value exceeds the preset deviation range, the pH value and stirring rate are updated through an iterative optimization algorithm to obtain the final coordinated change scheme.

[0058] In one specific embodiment, step S107 includes the following steps: If the deviation value exceeds the preset deviation range, the cause of the association between microbial imbalance and enzyme system disorder is analyzed based on the deviation value, and the direction and range of fine adjustment of pH value and stirring rate are determined. The particle swarm optimization algorithm is used to optimize the balance of the microbial community and the synergy of the enzyme system. The pH value and stirring rate are updated according to the determined fine-tuning direction and interval. After collecting updated parameters, the temperature, dissolved oxygen, pH, and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were measured, and the new deviation values ​​were calculated. Determine whether the new deviation value is within the preset deviation range. If not, repeat the above update and data collection calculation steps until the new deviation value meets the preset deviation range. Record the final pH value and stirring rate parameters, and combine them with the temperature and dissolved oxygen parameters in the optimal parameter combination to form the final coordinated change scheme.

[0059] Specifically, the preset deviation range is determined based on the stable operating threshold of the co-fermentation of Polygonatum sibiricum enzymes, with a value of 0~0.5. When the calculated comprehensive deviation value exceeds this range, it indicates that the system still fluctuates after the application of the optimal parameter combination, and the deviation needs to be corrected by fine-tuning the pH value and stirring rate. When analyzing the correlation between microbial imbalance and enzyme system imbalance based on the deviation value, it is necessary to combine the deviation contribution of each parameter dimension. The deviation contribution is obtained by multiplying the deviation value of a single parameter dimension by its weight. For example, in the comprehensive deviation value of 0.695, the contribution of temperature deviation is 0.03, dissolved oxygen is 0.045, pH value is 0.02, and stirring rate is 0.6. It can be seen that the stirring rate deviation is the main contributing factor. Combining the changes in microbial imbalance index and enzyme system imbalance index, it is judged that insufficient stirring rate leads to uneven dissolved oxygen distribution, which in turn causes microbial metabolic imbalance and enzyme activity fluctuation. Slight pH deviation exacerbates enzyme system imbalance. Based on this, the fine-tuning direction and range of pH value and stirring rate are determined. The current pH value is 6.6, with a small contribution to the deviation, so the fine-tuning direction is to slightly increase it, and the range is set to 6.6~6.8. The current stirring rate is 210 r / min, with a large contribution to the deviation, so the fine-tuning direction is to increase it, and the range is set to 210-240 r / min. This ensures that the fine-tuning range can correct the deviation without causing new parameter coupling conflicts.

[0060] When updating pH and stirring rate using particle swarm optimization, the optimization objectives are microbial community balance and enzyme synergy, and the objective function is set as follows: The microbial community imbalance correction rate and enzyme system synergistic maintenance rate were calculated using a prediction model trained on historical data. A larger objective function value indicates a better optimization effect. The algorithm parameters are set as follows: particle population size of 30, each particle corresponding to a combination of pH value and stirring rate, particle dimension of 2; maximum number of iterations of 50; and inertia weight. The initial value was 0.9, which decreased linearly to 0.4 with the number of iterations; cognitive coefficient Social coefficient These represent the learning ability of a particle to its own optimal position and the population's optimal position, respectively. When initializing the particle swarm, 30 sets of parameter combinations were randomly generated within the pH range of 6.6–6.8 and the stirring rate range of 210–240 r / min as initial particle positions. During iteration, each particle is assigned a position according to the formula... Update the velocity vector, where At the current speed, Current position This is the optimal position for the particle itself. The optimal position for the population and A random number between 0 and 1; then according to the formula Update the position vector to ensure the new position remains within the preset fine-tuning range. After each iteration, calculate the fitness value of each particle using the objective function, update the particle's own optimal position and the population's optimal position, until the maximum number of iterations is reached. Output the pH value and stirring rate corresponding to the population's optimal position. For example, after iteration, the pH value is 6.7 and the stirring rate is 225 r / min.

[0061] When collecting real-time data after parameter updates, the previous sensing equipment is used to collect temperature, dissolved oxygen, pH, and stirring rate data once per minute for one hour, totaling 60 sets of data. These data are then integrated into a new feedback dataset based on time series. The same root mean square error algorithm and weight allocation as in step S106 are used to calculate the new comprehensive deviation value. For example, if the new deviation value is 0.42, it falls within the preset deviation range of 0-0.5. If the new deviation value still exceeds the range, the iterative update, data collection, and deviation calculation steps of the particle swarm optimization algorithm are repeated, adjusting the algorithm parameters (e.g., increasing the iteration count to 80) and regenerating the fine-tuning parameter combination until the new deviation value meets the preset range. The final pH and stirring rate parameters are recorded. For example, if the final pH is determined to be 6.7 and the stirring rate 225 r / min, combined with the optimal parameter combination of 32℃ temperature and 7 mg / L dissolved oxygen, a final coordinated change scheme is formed. The scheme clearly indicates the specific values, control precision, and maintenance time of each parameter, providing operational guidance for the long-term stable operation of the fermentation system.

[0062] By accurately locating the causes of correlation through deviation contribution analysis, the direction and range of fine-tuning become more targeted. The application of particle swarm optimization algorithm enables the synergistic optimization of pH value and stirring rate, avoiding the chain reaction caused by fine-tuning a single parameter. The objective function setting takes into account both microbial community balance and enzyme synergy, solving the defects of multi-objective optimization conflict. The iterative update and deviation verification loop mechanism ensures the fine-tuning effect, so that the system deviation eventually converges to the stable range, constructing a complete closed loop of "adjustment-feedback-fine-tuning-verification". This overcomes the shortcomings of insufficient control precision and poor system stability in existing technologies, and further improves the product quality and process stability of Polygonatum sibiricum enzyme co-fermentation.

[0063] Please see Figure 2 , Figure 2 The graph shows the relationship between the out-of-bag error and the number of decision trees in a random forest model. It indicates that as the number of decision trees increases from 0 to 100, the out-of-bag error gradually decreases from approximately 0.023 and tends to stabilize (finally stabilizing at around 0.0187). This demonstrates that in determining the optimal parameters for co-fermentation of Polygonatum sibiricum enzymes, increasing the number of decision trees in the random forest model can effectively reduce the prediction error. However, once the number of decision trees reaches a certain scale (approximately 80 trees), the model's generalization ability no longer improves significantly.

[0064] Please see Figure 3 , Figure 3 The particle swarm optimization (PSO) iteration curve shows that as the number of iterations increases from 0 to 50, the best and average performance of the PSO gradually rises and tends to converge, while the worst performance improves simultaneously and the performance range gradually narrows. This demonstrates that in the iterative optimization of pH value and stirring rate, the PSO algorithm can approach the optimal solution through continuous iteration. After about 30 iterations, the performance becomes basically stable, verifying the effectiveness and convergence of the algorithm in parameter fine-tuning.

[0065] Please see Figure 4 , Figure 4 The comprehensive performance comparison chart shows that the comprehensive performance value of this method is approximately 0.9, while the comprehensive performance value of the traditional method is approximately 0.65. The significant difference between the two methods demonstrates that the intelligent control optimization method for co-fermentation of Polygonatum sibiricum enzymes, through its full-process design of real-time multi-parameter acquisition, intelligent model analysis, and closed-loop fine-tuning, outperforms traditional methods that rely on empirical rules or single-parameter regulation in terms of maintaining microbial balance, co-regulating enzyme systems, and ensuring fermentation stability. It can more efficiently solve the problem of multi-parameter coupling and interaction, and significantly improve the overall operating effect of the fermentation system.

[0066] Please see Figure 5 , Figure 5 The comparison chart of parameter control accuracy shows that the deviation curve of the traditional method fluctuates wildly, reaching a deviation of 1.2 in the early stage of fermentation. Although there are fluctuations thereafter, it always remains above 0.2, and even shows negative deviations at some times. The deviation curve of the present method is always close to the horizontal axis, with the fluctuation range controlled between -0.2 and 0.2, and maintained at an extremely low level throughout. This indicates that the present method, through a closed-loop control mechanism of "prediction-adjustment-feedback-fine-tuning", can correct deviations in temperature, dissolved oxygen, pH value and stirring rate in real time, accurately control the fluctuation of each parameter within the target range, and the control accuracy is significantly better than that of the traditional method. It effectively avoids the impact of large parameter fluctuations on the microbial community structure and enzyme activity, and provides a guarantee for the stable operation of the fermentation system.

[0067] Please see Figure 6 , Figure 6 The technical advantage comparison chart shows that in six dimensions—microbial community-enzyme synergy, multi-parameter coupling recognition, real-time intelligent regulation, prediction-feedback closed loop, energy efficiency, and product stability—the advantages of this method are higher than those of traditional methods. This indicates that this method overcomes the shortcomings of traditional methods in many aspects, constructs a comprehensive and efficient fermentation control system, and has multi-dimensional technical advantages.

[0068] Please see Figure 7 The intelligent control and optimization system for co-fermentation of Polygonatum sibiricum enzymes in this application embodiment is described below. The intelligent control and optimization system for co-fermentation of Polygonatum sibiricum enzymes includes: The data acquisition and analysis module collects real-time data on temperature, dissolved oxygen, pH, and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system. Based on the real-time data, it analyzes the microbial imbalance index and enzyme system disorder index and generates an overall description of the system's state. The influencing factor identification module is used to preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and identify the core influencing factors. The coupling interaction calculation module is used to input the core influencing factors into the pre-trained support vector machine model and calculate the parameter coupling interaction strength value. The candidate scheme screening module is used to simulate multi-parameter coordinated change scenarios when the parameter coupling interaction strength value exceeds the preset strength threshold, generate potential adjustment schemes, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes. The optimal parameter determination module is used to input the enzyme system co-maintenance prediction data into the pre-trained random forest model, evaluate the impact of candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination. The feedback deviation calculation module is used to adjust the operating parameters of the fermentation system according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the predicted data of enzyme system synergistic maintenance corresponding to the optimal parameter combination. The parameter optimization and adjustment module is used to update the pH value and stirring rate through an iterative optimization algorithm when the deviation value exceeds the preset deviation range, so as to obtain the final coordinated change scheme.

[0069] Through the synergistic cooperation of the above components, the system constructs a full-process intelligent control system for the co-fermentation of Polygonatum sibiricum and enzymes, encompassing "real-time monitoring, intelligent analysis, precise regulation, and closed-loop optimization." This system achieves end-to-end automated management from dynamic acquisition of multiple parameters of the fermentation system to precise maintenance of microbial community balance and enzyme synergy. The data acquisition and analysis module, serving as the foundation of the system, continuously captures real-time dynamics of temperature, dissolved oxygen, pH, and stirring rate using high-precision sensors. Combined with quantitative calculations of microbial imbalance and enzyme system dysregulation indicators, it generates a comprehensive description of the system's overall fermentation state, providing accurate data support for subsequent analysis and resolving the problem of ambiguous fermentation state assessments in traditional methods. The influencing factor identification module, through data preprocessing, feature extraction, and correlation analysis, accurately locates the core influencing factors leading to system instability from complex data, providing a clear direction for targeted regulation and overcoming the limitations of isolated single-parameter analysis. The coupled interaction calculation module uses a support vector machine model to quantify the strength of parameter coupling interactions, enabling early warning of strong coupling risks and avoiding [further risks]. The system avoids instability caused by parameter-free chain reactions. The candidate scheme screening module, after risk warning, simulates multi-parameter coordination scenarios and screens candidate schemes that meet enzyme synergy standards, providing high-quality alternatives for subsequent optimization. The optimal parameter determination module comprehensively evaluates the corrective effect of candidate schemes on microbial imbalance and the enzyme synergy maintenance ability through a random forest model, screening out the optimal parameter combination that takes into account multiple objectives to achieve synergistic optimization. The feedback deviation calculation module quantifies the deviation between real-time feedback data and predicted data, providing an objective evaluation basis for the control effect and ensuring the accuracy of the adjustment direction. The parameter optimization and adjustment module, based on deviation analysis, fine-tunes pH and stirring rate through iterative optimization algorithms, constructing a closed-loop control mechanism to ensure the long-term stable operation of the fermentation system. Each module is interconnected and works synergistically, achieving both accurate identification and quantification of multi-parameter coupling interactions and dynamic maintenance of microbial balance and enzyme synergy through a closed-loop mechanism of "prediction-adjustment-feedback-fine-tuning." This completely solves the defects of traditional methods, such as lagging regulation, lack of synergy, and poor stability, significantly improving the quality of Polygonatum fermentation products and process efficiency.

[0070] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the intelligent control and optimization method for co-fermentation of Polygonatum sibiricum enzyme.

[0071] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the methods and systems described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0072] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent control and optimization of co-fermentation of Polygonatum sibiricum and enzymes, characterized in that, Includes the following steps: Step S101: Collect real-time data on temperature, dissolved oxygen, pH and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system, analyze the microbial imbalance index and enzyme system disorder index based on the real-time data, and generate an overall state description of the system. Step S102: Preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and determine the core influencing factors; Step S103: Input the core influencing factors into the pre-trained support vector machine model and calculate the parameter coupling interaction strength value; Step S104: If the parameter coupling interaction strength value exceeds the preset strength threshold, simulate the multi-parameter coordinated change scenario to generate potential adjustment schemes, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes. Step S105: Input the enzyme system co-maintenance prediction data into a pre-trained random forest model, evaluate the impact of the candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination; Step S106: Adjust the fermentation system operating parameters according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the predicted data of enzyme system synergistic maintenance corresponding to the optimal parameter combination. Step S107: If the deviation value exceeds the preset deviation range, the pH value and stirring rate are updated through an iterative optimization algorithm to obtain the final coordinated change scheme.

2. The method according to claim 1, characterized in that, Step S101 includes: Real-time data on temperature, dissolved oxygen, pH, and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were collected and integrated to form a dynamic dataset. The temperature change rate and dissolved oxygen fluctuation amplitude are calculated based on the dynamic dataset, and the ratio of the temperature change rate to the dissolved oxygen fluctuation amplitude is used as an indicator of microbial imbalance. The pH deviation and stirring rate adjustment frequency are calculated based on the dynamic dataset, and the product of the pH deviation and the stirring rate adjustment frequency is used as an indicator of enzyme system imbalance. Based on the microbial community imbalance index and the enzyme system disorder index, a description of the overall system state characterizing the operation of the fermentation system is generated.

3. The method according to claim 2, characterized in that, Step S102 includes: The data in the overall state description of the system are preprocessed to obtain standardized state data; Based on the standardized state data, principal component analysis was used to extract the feature vectors of microbial community imbalance, enzyme system disorder, and environmental parameters to construct a feature dataset. For the aforementioned feature dataset, the correlation coefficient between the microbial imbalance feature vector and the enzyme system imbalance feature vector is calculated, and high-coefficient indicators with an absolute value of correlation coefficient greater than a preset correlation threshold are selected. A decision tree model is constructed based on the high-coefficient indicators, and the decision tree model is pruned to determine the core influencing factors.

4. The method according to claim 1, characterized in that, Step S103 includes: The core influencing factors are converted into feature factor vectors and input into a pre-trained support vector machine model. The support vector machine model is used to classify the inhibition demand of dominant bacteria and the activation potential of weak bacteria in the Polygonatum fermentation system, and the classification probability vector is output. Based on the classification probability vector, calculate the parameter interaction coefficients between temperature, dissolved oxygen, pH value and stirring rate corresponding to each core influencing factor. The parameter coupling interaction strength value is obtained by weighted summation based on the weight proportion of the classification probability vector and the parameter interaction coefficient.

5. The method according to claim 1, characterized in that, Step S104 includes: If the parameter coupling interaction strength value exceeds the preset strength threshold, the temperature, dissolved oxygen, pH value and stirring rate in the fermentation system are used as the control objects to simulate multiple sets of multi-parameter coordinated change scenarios and generate multiple potential adjustment schemes. The effects of each potential adjustment scheme on enzyme synergy in the Polygonatum sibiricum enzyme co-fermentation system were analyzed, and the degree of enzyme activity balance and the potential for correcting enzyme imbalance were quantified. Based on the impact analysis results, predictive data for enzyme system synergistic maintenance corresponding to each potential adjustment scheme are generated. Potential adjustment schemes that meet the preset enzyme system synergy criteria in the enzyme system synergy maintenance prediction data are selected as candidate schemes.

6. The method according to claim 1, characterized in that, Step S105 includes: The enzyme system synergistic maintenance prediction data corresponding to the candidate schemes are correlated and integrated with the microbial community imbalance indicators to form the model input dataset; The model input dataset is input into a pre-trained random forest model, and the influence coefficient of community imbalance correction corresponding to each candidate scheme is output. Based on the community imbalance correction influence coefficient, and combined with the enzyme system synergistic maintenance prediction data corresponding to the candidate schemes, the comprehensive optimization value of each candidate scheme is quantified. The optimal parameter combination is determined by selecting the candidate scheme with the highest comprehensive optimization value corresponding to the combination of temperature, dissolved oxygen, pH value, and stirring rate.

7. The method according to claim 1, characterized in that, Step S106 includes: Based on the optimal parameter combination, the temperature, dissolved oxygen, pH value and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were adjusted. Real-time feedback data on temperature, dissolved oxygen, pH, and stirring rate in the fermentation system after parameter adjustment are continuously collected and integrated into a feedback dataset according to the time series. The feedback dataset is matched with the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination according to the corresponding parameter dimensions; Based on the matching results, the deviation between the real-time feedback data and the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination is calculated.

8. The method according to claim 1, characterized in that, Step S107 includes: If the deviation value exceeds the preset deviation range, the cause of the association between the bacterial community imbalance and enzyme system disorder is analyzed based on the deviation value, and the direction and range of fine adjustment of pH value and stirring rate are determined. The particle swarm optimization algorithm is used to optimize the balance of the microbial community and the synergy of the enzyme system. The pH value and stirring rate are updated according to the determined fine-tuning direction and interval. After collecting updated parameters, the temperature, dissolved oxygen, pH, and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system were measured, and the new deviation values ​​were calculated. Determine whether the new deviation value is within the preset deviation range. If not, repeat the above update and data collection calculation steps until the new deviation value meets the preset deviation range. Record the final pH value and stirring rate parameters, and combine them with the temperature and dissolved oxygen parameters in the optimal parameter combination to form the final coordinated change scheme.

9. A smart control and optimization system for co-fermentation of Polygonatum sibiricum enzymes, used to implement the smart control and optimization method for co-fermentation of Polygonatum sibiricum enzymes as described in any one of claims 1 to 8, characterized in that, The aforementioned intelligent control and optimization system for co-fermentation of Polygonatum sibiricum and enzymes includes: The data acquisition and analysis module collects real-time data on temperature, dissolved oxygen, pH value and stirring rate in the Polygonatum sibiricum enzyme co-fermentation system. Based on the real-time data, it analyzes the microbial imbalance index and enzyme system disorder index and generates an overall description of the system's state. The influencing factor determination module is used to preprocess the overall state description of the system and extract feature vectors, analyze the correlation between microbial imbalance and enzyme system disorder, and determine the core influencing factors. The coupling interaction calculation module is used to input the core influencing factors into a pre-trained support vector machine model and calculate the parameter coupling interaction strength value. The candidate scheme screening module is used to simulate a multi-parameter coordinated change scenario to generate potential adjustment schemes when the parameter coupling interaction strength value exceeds a preset strength threshold, analyze the impact of potential adjustment schemes on enzyme synergy and generate corresponding enzyme synergy maintenance prediction data, and screen to obtain candidate schemes. The optimal parameter determination module is used to input the enzyme system co-maintenance prediction data into a pre-trained random forest model, evaluate the impact of the candidate schemes on the correction of microbial imbalance, and determine the optimal parameter combination. The feedback deviation calculation module is used to adjust the operating parameters of the fermentation system according to the optimal parameter combination, collect real-time feedback data, and calculate the deviation value between the enzyme system synergistic maintenance prediction data corresponding to the optimal parameter combination. The parameter optimization and adjustment module is used to update the pH value and stirring rate through an iterative optimization algorithm when the deviation value exceeds the preset deviation range, so as to obtain the final coordinated change scheme.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor executes the intelligent control and optimization method for co-fermentation of Polygonatum odoratum enzyme as described in any one of claims 1 to 8.

Images