A multi-parameter adaptive control fermentation monitoring system

Abstract

This invention belongs to the field of fermentation monitoring technology. It discloses a multi-parameter adaptive control fermentation monitoring system. The system includes: a fermentation risk prediction module, a cross-batch adaptive learning module, a multi-timescale decision-making module, a multi-objective collaborative optimization module, and a risk quantification and control module. The system processes feature vectors to obtain fermentation risk prediction values ​​and generates weight update reports. Based on the fermentation risk prediction values, it processes fermentation cycle stages to obtain multi-timescale reports. Based on the risk prediction values, it optimizes multi-objective requirements to obtain collaborative optimization reports. The system also performs hierarchical processing of risk prediction values ​​and analyzes the processing results to obtain risk response decision reports. Overall, this invention has significant advantages such as high accuracy in fermentation risk quantification prediction, strong adaptive learning ability, and good decision support effect.

Description

Technical Field

[0001] This invention relates to the field of fermentation monitoring technology, and more specifically, to a fermentation monitoring system with multi-parameter adaptive control. Background Technology

[0002] Fermentation is the core step in alcoholic beverage production. During fermentation, microorganisms decompose raw materials to produce flavor compounds such as esters and alcohols. For example, the main aroma compounds in light-aroma baijiu and strong-aroma baijiu, such as ethyl acetate, originate from fermentation. Fermentation monitoring is an indispensable and important step in alcoholic beverage production. It is mainly used to collect and analyze key data in real time during the fermentation process to ensure that the microorganisms are in the optimal state of growth and product synthesis, and ultimately achieve the production goals of high yield, stability, and low consumption.

[0003] However, traditional fermentation detection systems mostly suffer from the following shortcomings: First, traditional systems are poor at quantifying and responding to risks during monitoring. Because they typically only compare single data points (such as temperature or pH) with fixed thresholds before issuing alarms, the systems lack the ability to comprehensively quantify and predict risks under the coupling of multiple parameters. Second, the models used for prediction or calculation in traditional systems are usually fixed and unchanging, making it difficult to effectively utilize historical batch data for continuous optimization and adaptive learning of the models, thus reducing the practicality and economy of the system. Third, at the control level, the control strategies of traditional systems are mostly rigid, lacking the ability to make decision-making suggestions across multiple time scales and multiple objectives, making the system highly dependent on human judgment, thereby reducing the applicability of the system. Overall, how to effectively solve the problems of insufficient risk quantification, rigid operating modes, and poor decision-making support capabilities of traditional systems has become the main obstacle that needs to be addressed.

[0004] In view of this, the present invention proposes a fermentation monitoring system with multi-parameter adaptive control to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution, including:

[0006] The fermentation data acquisition module is used to collect fermentation datasets, which include pH data, dissolved oxygen data, fermentation temperature data, cell concentration data, and ethanol concentration data.

[0007] The environmental data acquisition module is used to collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data.

[0008] The data feature extraction module is used to preprocess the fermentation dataset and the environmental dataset, and to perform data interaction processing to obtain feature vectors;

[0009] Furthermore, the steps of preprocessing the fermentation dataset and the environmental dataset, and performing data interaction processing, include:

[0010] Q1: Clean the data by removing outliers and normalize all sub-data items in the basic dataset to the range of [0, 1] according to the normalization formula;

[0011] Q2: Calculate pH value data and fermentation temperature data The product of these two elements yields the first interactive feature data. ;

[0012] Q3: Calculate dissolved oxygen data and cell concentration data The product of these two elements yields the second interactive feature data. ;

[0013] Q4: Calculate the contamination index data and bubble level data The product of these factors yields the third interactive feature data. ;

[0014] Q5: Calculate cell concentration data Ethanol concentration data The product of these elements yields the fourth interactive feature data. ;

[0015] Q6: Package the first interaction feature data, the second interaction feature data, the third interaction feature data, and the tank pressure data to obtain the feature vector;

[0016] The fermentation risk prediction module is used to process the feature vector to obtain the fermentation risk prediction value;

[0017] Furthermore, the steps for processing the feature vectors include:

[0018] Step 1: Obtain a set of historical feature vectors stored in the database, and compare them with the current time based on the timestamp. Group the comparison results from smallest to largest and label them accordingly. The labeling results are L1, L2, L3, ..., Ln. Use the labeling results as the sample set.

[0019] Step 2: Divide the sample set into 70% training set, 15% test set and 15% validation set, and build a fermentation risk prediction model based on the sample set.

[0020] Step 3: Based on the historical feature vectors in the training set, denote the feature vector at the current time as the current time step vector. And define the feature vectors of the first two time steps, denoted as , respectively. and ;

[0021] Step 4: Define an extended vector based on the current time step vector, the previous time step vector, and the vector from the second-to-last time step in Step 3. The expression for the extended vector is: ,in, , The interaction term includes the result of the interaction multiplication of the four feature data in the feature vector;

[0022] Step 5: Calculate the predicted fermentation risk value based on the extended vector from Step 4. The specific formula for calculation is:

[0023] ;

[0024] in, It is the Sigmoid activation function. For bias terms, Let be the dimension of the extended vector. For the first Dimensional weighting factors;

[0025] Step Six: Output the fermentation risk prediction value to the multi-timescale decision-making module, the multi-objective collaborative optimization module, and the risk quantification and control module;

[0026] A cross-batch adaptive learning module is used to optimize the fermentation risk prediction module based on historical batch success and failure data, and to obtain a weight update report.

[0027] Furthermore, the steps for optimizing the fermentation risk prediction module based on historical batch success and failure data include:

[0028] W1: Retrieves a set of historical fermentation risk prediction values ​​and historical actual risk values ​​stored in the database;

[0029] W2: The loss function is obtained by calculating the square of the historical true risk value minus the historical fermentation risk prediction value within the same group;

[0030] W3: Update the weight factors in step five of the fermentation risk prediction module based on the gradient descent method to obtain new weight factors. The specific calculation formula for the update is:

[0031] ;

[0032] in, For learning rate, For loss function, These are partial derivatives;

[0033] W4: Output the new weighting factor to the fermentation risk prediction module and use it for the next calculation of the fermentation risk prediction model;

[0034] The multi-timescale decision module is used to process the fermentation cycle stages based on the fermentation risk prediction value and obtain multi-timescale reports.

[0035] Furthermore, the steps for processing different stages of the fermentation cycle based on predicted fermentation risk values ​​include:

[0036] E1: Based on the changing trends of cell concentration data, the fermentation phase is divided into the lag phase, log phase, stationary phase, and death phase;

[0037] E2: When the predicted fermentation risk value is greater than or equal to the high fermentation threshold, an adjustment instruction is generated;

[0038] The adjustment command contains a set of characters representing the adjustment of the fermentation temperature data to the target temperature;

[0039] E3: When the fermentation stage is logarithmic, if the predicted fermentation risk is less than the first adjustment threshold, an increase in feeding report is generated; if the predicted fermentation risk is greater than the second adjustment threshold, a decrease in feeding report is generated.

[0040] The replenishment report should include a description of the need to increase the replenishment rate.

[0041] A reduced replenishment report should include an explanation of the need to reduce the replenishment rate.

[0042] E4: Packaging adjustment instructions, adding replenishment reports and reducing replenishment reports, resulting in multi-timescale reports;

[0043] The multi-objective collaborative optimization module is used to optimize multi-objective requirements based on risk prediction values ​​and generate a collaborative optimization report;

[0044] Furthermore, the steps for optimizing multi-objective requirements based on risk prediction values ​​include:

[0045] R1: Multi-objective requirements include maximizing output, minimizing energy consumption, and minimizing byproduct inhibition;

[0046] R2: Variables for optimizing multi-objective requirements include feed rate, stirring rate, and fermentation temperature setpoint;

[0047] R3: Based on maximizing output as the primary objective, an optimization problem is established;

[0048] R4: Solve the optimization problem based on the genetic algorithm to obtain the optimal feeding rate, optimal stirring rate and optimal fermentation temperature setpoints respectively;

[0049] R5: The optimal feeding rate, optimal stirring rate, and optimal fermentation temperature settings for packaging are used to obtain a collaborative optimization report;

[0050] The risk quantification and management module is used to classify and process risk prediction values, analyze the processing results, and obtain a risk response decision report.

[0051] Furthermore, methods for categorizing risk prediction values ​​and analyzing the results include:

[0052] Based on the risk threshold interval (Y1, Y2);

[0053] When the predicted risk value is less than Y1, a low-risk report is generated; when the predicted risk value is greater than or equal to Y1 and less than Y2, a medium-risk report and parameter adjustment instructions are generated; when the predicted risk value is greater than or equal to Y2, a high-risk report and stop instructions are generated.

[0054] The low-risk report indicates that the predicted risk of fermentation is low, and staff should maintain normal monitoring without intervention.

[0055] The medium-risk report includes parameter adjustment instructions and explanations predicting the level of fermentation risk. Staff should pay attention to system adjustments and go to the fermentation site to confirm.

[0056] The high-risk report includes a stop order and a statement indicating a high level of predicted fermentation risk. Staff are requested to immediately enter the site to conduct aseptic sampling after confirming that fermentation has stopped, and to activate the fermenter cleaning mechanism.

[0057] The parameter adjustment command contains a set of characters representing a reduction in fermentation temperature data to ninety percent;

[0058] The stop command contains a set of characters that represent stopping the fermenter from running;

[0059] Package low-risk, medium-risk, and high-risk reports to obtain a risk response decision report;

[0060] The system data management module is used to store system datasets, display multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and process multi-timescale reports and risk response decision reports;

[0061] Furthermore, the methods for processing multi-timescale reports and risk response decision-making reports include:

[0062] Monitor multi-timescale reports and risk response decision reports for any adjustment instructions, parameter adjustment instructions, or stop instructions;

[0063] When there is an adjustment command, parameter adjustment command, or stop command, the adjustment command, parameter adjustment command, or stop command is converted into a PLC-readable signal through the OPC UA protocol and sent to the intelligent controller.

[0064] The system dataset includes fermentation dataset, environmental dataset, feature vectors, fermentation risk prediction values, multi-timescale reports, collaborative optimization reports, and risk response decision reports;

[0065] Further, S1: Collect fermentation dataset, which includes pH data, dissolved oxygen data, fermentation temperature data, cell concentration data, and ethanol concentration data;

[0066] S2: Collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data;

[0067] S3: Preprocess the fermentation dataset and the environmental dataset, and perform data interaction processing to obtain feature vectors;

[0068] S4: Process the feature vector to obtain the fermentation risk prediction value;

[0069] S5: Optimize the fermentation risk prediction module based on historical batch success and failure data to obtain a weight update report;

[0070] S6: Process the fermentation cycle stages based on the fermentation risk prediction values ​​to obtain multi-timescale reports;

[0071] S7: Optimize multi-objective requirements based on risk prediction values ​​to obtain a collaborative optimization report;

[0072] S8: Classify the risk prediction values ​​and analyze the results to obtain a risk response decision report;

[0073] S9: Stores the system dataset, displays multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and processes the risk response decision reports.

[0074] The technical effects and advantages of the multi-parameter adaptive control fermentation monitoring system of the present invention are as follows:

[0075] This invention collects a fermentation dataset, including pH, dissolved oxygen, fermentation temperature, cell concentration, and ethanol concentration data, and an environmental dataset, including microbial content index, foam level, and tank pressure data. The fermentation and environmental datasets are preprocessed and interacted to obtain feature vectors. These feature vectors are then processed to obtain fermentation risk prediction values. The fermentation risk prediction module is optimized based on historical batch success and failure data to generate a weight update report. The fermentation cycle stages are processed according to the predicted fermentation risk values ​​to generate multi-timescale reports. Multiple objective requirements are optimized based on the risk prediction values ​​to generate a collaborative optimization report. The risk prediction values ​​are then graded and analyzed to generate a risk response decision report. The system stores a unified dataset and displays multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel. It processes these risk response decision reports, enabling the system to accurately quantify and predict fermentation risks based on multiple parameters, significantly improving its foresight in predicting fermentation risks. Furthermore, the invention utilizes a cross-batch adaptive learning module to effectively update the prediction model using historical data, greatly enhancing prediction accuracy and cost-effectiveness. Simultaneously, by generating multi-timescale and multi-dimensional auxiliary decision reports, the system achieves multi-dimensional decision support, significantly reducing the reliance on human experience in traditional systems. Overall, this invention possesses significant advantages: high accuracy in quantitative prediction of fermentation risks, strong adaptive learning capabilities, and effective decision support. Attached Figure Description

[0076] Figure 1 This is a schematic diagram of a multi-parameter adaptive control fermentation monitoring system according to the present invention;

[0077] Figure 2 This is a schematic diagram of a multi-parameter adaptive control fermentation monitoring method according to the present invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0080] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0081] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0082] In practice, the server-side equipment deployed in a multi-parameter adaptive control fermentation monitoring system may consist of one or more devices. This multi-parameter adaptive control fermentation monitoring system can be implemented as a business instance, a virtual machine, or a hardware device. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node, providing multi-parameter adaptive control fermentation monitoring to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various user terminals. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more devices configured to provide multi-parameter adaptive control fermentation monitoring to various user terminals.

[0083] In terms of implementation, the fermentation monitoring system with multi-parameter adaptive control and the user terminal are mutually compatible. That is, if the fermentation monitoring system with multi-parameter adaptive control is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the fermentation monitoring system with multi-parameter adaptive control is implemented as a website, then the user terminal is implemented as a webpage; or if the fermentation monitoring system with multi-parameter adaptive control is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0084] like Figure 1 The figure shown is a system architecture diagram of a fermentation monitoring system with multi-parameter adaptive control provided in an embodiment of the present invention.

[0085] The multi-parameter adaptive control fermentation monitoring system of this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the multi-parameter adaptive control fermentation monitoring system may include a fermentation data acquisition module, an environmental data acquisition module, a data feature extraction module, a fermentation risk prediction module, a cross-batch adaptive learning module, a multi-timescale decision-making module, a multi-objective collaborative optimization module, a risk quantification and control module, and a system data management module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0086] In this embodiment of the invention, in the fermentation monitoring system with multi-parameter adaptive control, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as a module connecting to multiple modules of another type and providing corresponding services to those connected modules. For example, the sharing and evaluation module can call the same information acquisition module to obtain information collected by that module. Based on the above characteristics, in the fermentation monitoring system with multi-parameter adaptive control provided in this embodiment of the invention, the applicable scope of the fermentation monitoring system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the fermentation monitoring system with multi-parameter adaptive control. In practical applications, the above modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0087] Example 1

[0088] Please see Figure 1 As shown in this embodiment, a multi-parameter adaptive control fermentation monitoring system includes:

[0089] The fermentation data acquisition module is used to acquire fermentation datasets, which include pH data, dissolved oxygen data, fermentation temperature data, cell concentration data, and ethanol concentration data.

[0090] It should be explained that the pH sensor collects real-time pH values ​​within the designated fermenter to obtain pH data; the dissolved oxygen sensor collects the percentage of dissolved oxygen concentration within the designated fermenter to obtain dissolved oxygen data; the temperature sensor collects real-time temperature values ​​within the designated fermenter to obtain fermentation temperature data; the Merck optical density sensor collects cell concentration values ​​of the sample within the fermenter to obtain cell concentration data; and the ethanol gas sensor collects the ethanol concentration of the fermentation exhaust gas within the designated fermenter and calculates the ethanol concentration in the fermentation broth to obtain ethanol concentration data.

[0091] The environmental data acquisition module is used to collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data.

[0092] It should be explained that the total amount of microorganisms in a designated fermenter is collected using a bioluminescence rapid detector to obtain the miscellaneous bacteria content index data; the height of foam generation in a designated fermenter is collected using a conductive foam probe to obtain foam level data; and the pressure value in a designated fermenter is collected using a pressure transmitter to obtain tank pressure data.

[0093] The data feature extraction module is used to preprocess the fermentation dataset and the environmental dataset, and to perform data interaction processing to obtain feature vectors.

[0094] Further steps include preprocessing the fermentation dataset and the environmental dataset, and performing data interaction processing:

[0095] Q1: Clean the data by removing outliers and normalize all sub-data items in the basic dataset to the range of [0, 1] according to the normalization formula;

[0096] It should be explained that removing outliers refers to, for example, negative values ​​in foam level data; the base dataset includes the fermentation dataset and the environment dataset; the specific expression of the normalization formula is: ,in Normalized value Any sub-data item of the basic data, This represents the historical maximum value of any given sub-data item. This represents the historical minimum value of any given sub-data item; normalization is used to eliminate the dimensions of all sub-data items in the base dataset.

[0097] Q2: Calculate pH value data and fermentation temperature data The product of these two elements yields the first interactive feature data. ;

[0098] Q3: Calculate dissolved oxygen data and cell concentration data The product of these two elements yields the second interactive feature data. ;

[0099] Q4: Calculate the contamination index data and bubble level data The product of these factors yields the third interactive feature data. ;

[0100] Q5: Calculate cell concentration data Ethanol concentration data The product of these elements yields the fourth interactive feature data. ;

[0101] Q6: Package the first interaction feature data, the second interaction feature data, the third interaction feature data, and the tank pressure data to obtain the feature vector;

[0102] It should be explained that the sub-data items in the basic data involved in steps Q2 to Q4 and all subsequent modules are data values ​​that have been cleaned and normalized.

[0103] The fermentation risk prediction module is used to process the feature vector to obtain the fermentation risk prediction value.

[0104] Further steps for processing the feature vectors include:

[0105] Step 1: Obtain a set of historical feature vectors stored in the database, and compare them with the current time based on the timestamp. Group the comparison results from smallest to largest and label them accordingly. The labeling results are L1, L2, L3, ..., Ln. Use the labeling results as the sample set.

[0106] Step 2: Divide the sample set into 70% training set, 15% test set and 15% validation set, and build a fermentation risk prediction model based on the sample set.

[0107] Step 3: Based on the historical feature vectors in the training set, denote the feature vector at the current time as the current time step vector. And define the feature vectors of the first two time steps, denoted as , respectively. and ;

[0108] Step 4: Define an extended vector based on the current time step vector, the previous time step vector, and the vector from the second-to-last time step in Step 3. The expression for the extended vector is: ,in, , The interaction term includes the result of the interaction multiplication of the four feature data in the feature vector;

[0109] Step 5: Calculate the predicted fermentation risk value based on the extended vector from Step 4. The specific formula for calculation is:

[0110] ;

[0111] in, It is the Sigmoid activation function. For bias terms, Let be the dimension of the extended vector. For the first Dimensional weighting factors;

[0112] It should be explained that the Sigmoid activation function is used to constrain the output value within the immediately following brackets to the range (0,1);

[0113] Step Six: Output the fermentation risk prediction value to the multi-timescale decision-making module, the multi-objective collaborative optimization module, and the risk quantification and control module;

[0114] The cross-batch adaptive learning module is used to optimize the fermentation risk prediction module based on the success and failure data of historical batches, and to obtain a weight update report.

[0115] Furthermore, the steps for optimizing the fermentation risk prediction module based on historical batch success and failure data include:

[0116] W1: Retrieves a set of historical fermentation risk prediction values ​​and historical actual risk values ​​stored in the database;

[0117] W2: The loss function is obtained by calculating the square of the historical true risk value minus the historical fermentation risk prediction value within the same group;

[0118] W3: Update the weight factors in step five of the fermentation risk prediction module based on the gradient descent method to obtain new weight factors. The specific calculation formula for the update is:

[0119] ;

[0120] in, For learning rate, For loss function, These are partial derivatives;

[0121] W4: Output the new weighting factor to the fermentation risk prediction module and use it for the next calculation of the fermentation risk prediction model;

[0122] The multi-timescale decision module is used to process the fermentation cycle stages based on the fermentation risk prediction value and obtain a multi-timescale report.

[0123] Furthermore, the steps for processing different stages of the fermentation cycle based on predicted fermentation risk values ​​include:

[0124] E1: Based on the changing trends of cell concentration data, the fermentation phase is divided into the lag phase, log phase, stationary phase, and death phase;

[0125] It should be noted that the cell concentration data in this step are the raw data without normalization;

[0126] E2: When the predicted fermentation risk value is greater than or equal to the high fermentation threshold, an adjustment instruction is generated;

[0127] The adjustment command contains a set of characters representing the adjustment of the fermentation temperature data to the target temperature;

[0128] It should be noted that the fermentation temperature data in this step is the raw data without normalization; the high fermentation threshold is manually set and entered into the system; the target temperature is calculated by substituting into the formula: We obtained, among which, This is the gain coefficient. High fermentation threshold;

[0129] E3: When the fermentation stage is logarithmic, if the predicted fermentation risk is less than the first adjustment threshold, an increase in feeding report is generated; if the predicted fermentation risk is greater than the second adjustment threshold, a decrease in feeding report is generated.

[0130] The replenishment report should include a description of the need to increase the replenishment rate.

[0131] A reduced replenishment report should include an explanation of the need to reduce the replenishment rate.

[0132] E4: Packaging adjustment instructions, adding replenishment reports and reducing replenishment reports, resulting in multi-timescale reports;

[0133] The multi-objective collaborative optimization module is used to optimize multi-objective requirements based on risk prediction values ​​and obtain a collaborative optimization report;

[0134] Furthermore, the steps for optimizing multi-objective requirements based on risk prediction values ​​include:

[0135] R1: Multi-objective requirements include maximizing output, minimizing energy consumption, and minimizing byproduct inhibition;

[0136] R2: Variables for optimizing multi-objective requirements include feed rate, stirring rate, and fermentation temperature setpoint;

[0137] R3: Based on maximizing output as the primary objective, an optimization problem is established;

[0138] It should be explained that the optimization problem refers to finding a set of variables that maximizes the value of the primary objective requirement;

[0139] R4: Solve the optimization problem based on the genetic algorithm to obtain the optimal feeding rate, optimal stirring rate and optimal fermentation temperature setpoints respectively;

[0140] R5: The optimal feeding rate, optimal stirring rate, and optimal fermentation temperature settings for packaging are used to obtain a collaborative optimization report;

[0141] The risk quantification and control module is used to classify and process the risk prediction values, and analyze the processing results to obtain a risk response decision report.

[0142] Furthermore, methods for categorizing risk prediction values ​​and analyzing the results include:

[0143] Based on the risk threshold interval (Y1, Y2);

[0144] It should be explained that the risk threshold range is manually set and entered into the system;

[0145] When the predicted risk value is less than Y1, a low-risk report is generated; when the predicted risk value is greater than or equal to Y1 and less than Y2, a medium-risk report and parameter adjustment instructions are generated; when the predicted risk value is greater than or equal to Y2, a high-risk report and stop instructions are generated.

[0146] The low-risk report indicates that the predicted risk of fermentation is low, and staff should maintain normal monitoring without intervention.

[0147] The medium-risk report includes parameter adjustment instructions and explanations predicting the level of fermentation risk. Staff should pay attention to system adjustments and go to the fermentation site to confirm.

[0148] The high-risk report includes a stop order and a statement indicating a high level of predicted fermentation risk. Staff are requested to immediately enter the site to conduct aseptic sampling after confirming that fermentation has stopped, and to activate the fermenter cleaning mechanism.

[0149] The parameter adjustment command contains a set of characters representing a reduction in fermentation temperature data to ninety percent;

[0150] It should be noted that the fermentation temperature data in this module is the raw data without normalization;

[0151] The stop command contains a set of characters that represent stopping the fermenter from running;

[0152] Package low-risk, medium-risk, and high-risk reports to obtain a risk response decision report;

[0153] The system data management module is used to store the system dataset, display multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and process the multi-timescale reports and risk response decision reports.

[0154] Furthermore, methods for processing multi-timescale reports and risk response decision-making reports include:

[0155] Monitor multi-timescale reports and risk response decision reports for any adjustment instructions, parameter adjustment instructions, or stop instructions;

[0156] When there is an adjustment command, parameter adjustment command, or stop command, the adjustment command, parameter adjustment command, or stop command is converted into a PLC-readable signal through the OPC UA protocol and sent to the intelligent controller.

[0157] The system dataset includes fermentation dataset, environmental dataset, feature vectors, fermentation risk prediction values, multi-timescale reports, collaborative optimization reports, and risk response decision reports;

[0158] In this embodiment, the beneficial effects are achieved by collecting a fermentation dataset, including pH, dissolved oxygen, fermentation temperature, cell concentration, and ethanol concentration data, and an environmental dataset, including microbial content index, foam level, and tank pressure data. Both the fermentation and environmental datasets are preprocessed and interacted with to obtain feature vectors. These feature vectors are then processed to obtain fermentation risk prediction values. The fermentation risk prediction module is optimized based on historical batch success and failure data, resulting in a weight update report. The fermentation cycle stages are processed according to the predicted fermentation risk values ​​to obtain multi-timescale reports. Multiple objective requirements are optimized based on the risk prediction values ​​to obtain a collaborative optimization report. Finally, the risk prediction values ​​are graded, and the processing results are analyzed to obtain a risk response decision report. The system stores the dataset and displays multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel. It processes these risk response decision reports, enabling the system to accurately quantify and predict fermentation risks based on multiple parameters, significantly improving its foresight in predicting fermentation risks. Furthermore, the invention utilizes a cross-batch adaptive learning module to effectively update the prediction model using historical data, greatly enhancing prediction accuracy and cost-effectiveness. Simultaneously, by generating multi-timescale and multi-dimensional auxiliary decision reports, the system achieves multi-dimensional decision support, significantly reducing the reliance on human experience in traditional systems. Overall, this invention has significant advantages: high accuracy in quantitative prediction of fermentation risks, strong adaptive learning capabilities, and effective decision support.

[0159] Example 2

[0160] Please see Figure 2 As shown, the parts not described in detail in this embodiment are described in Embodiment 1. A fermentation monitoring method with multi-parameter adaptive control is provided. The method includes: S1: collecting fermentation dataset, which includes pH data, dissolved oxygen data, fermentation temperature data, cell concentration data and ethanol concentration data;

[0161] S2: Collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data;

[0162] S3: Preprocess the fermentation dataset and the environmental dataset, and perform data interaction processing to obtain feature vectors;

[0163] S4: Process the feature vector to obtain the fermentation risk prediction value;

[0164] S5: Optimize the fermentation risk prediction module based on historical batch success and failure data to obtain a weight update report;

[0165] S6: Process the fermentation cycle stages based on the fermentation risk prediction values ​​to obtain multi-timescale reports;

[0166] S7: Optimize multi-objective requirements based on risk prediction values ​​to obtain a collaborative optimization report;

[0167] S8: Classify the risk prediction values ​​and analyze the results to obtain a risk response decision report;

[0168] S9: Stores the system dataset, displays multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and processes the risk response decision reports.

[0169] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the present invention.

Claims

1. A fermentation monitoring system with multi-parameter adaptive control, characterized in that, The system includes: a fermentation data acquisition module, an environmental data acquisition module, a data feature extraction module, a fermentation risk prediction module, a cross-batch adaptive learning module, a multi-timescale decision-making module, a multi-objective collaborative optimization module, and a risk quantification and control module, wherein: The fermentation data acquisition module is used to acquire fermentation datasets, which include pH data, fermentation temperature data, cell concentration data, and ethanol concentration data. The environmental data acquisition module is used to collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data. The data feature extraction module is used to preprocess the fermentation dataset and the environmental dataset, and to perform data interaction processing to obtain feature vectors. The fermentation risk prediction module is used to process the feature vector to obtain the fermentation risk prediction value. The steps for processing feature vectors include: Step 1: Obtain a set of historical feature vectors stored in the database, and compare them with the current time based on the timestamp. Group the comparison results from smallest to largest and label them accordingly. The labeling results are L1, L2, L3, ..., Ln. Use the labeling results as the sample set. Step 2: Divide the sample set into 70% training set, 15% test set and 15% validation set, and build a fermentation risk prediction model based on the sample set. Step 3: Based on the historical feature vectors in the training set, denote the feature vector of the current time as the current time step vector, and define the feature vectors of the previous two time steps; Step 4: Define the extended vector based on the current time step vector, the previous time step vector, and the two previous time step vectors from Step 3; Step 5: Calculate the fermentation risk prediction value based on the extended vector from Step 4; Step Six: Output the fermentation risk prediction value to the multi-timescale decision-making module, the multi-objective collaborative optimization module, and the risk quantification and control module; The cross-batch adaptive learning module is used to optimize the fermentation risk prediction module based on the success and failure data of historical batches, and to obtain a weight update report. The multi-timescale decision module is used to process the fermentation cycle stages based on risk prediction values ​​and obtain multi-timescale reports. The multi-objective collaborative optimization module is used to optimize multi-objective requirements based on risk prediction values ​​and obtain a collaborative optimization report; The risk quantification and control module is used to classify and process the risk prediction values, and analyze the processing results to obtain a risk response decision report. The system also includes a system data management module, wherein: The system data management module is used to store system datasets, display multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and process the risk response decision reports.

2. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, The steps for preprocessing the fermentation dataset and the environmental dataset, and for performing data interaction processing, include: Q1: Clean the data by removing outliers and normalize all sub-data items in the basic dataset to the range of [0, 1] according to the normalization formula; Q2: Calculate the product of the pH data and the fermentation temperature data to obtain the first interactive feature data. ; Q3: Calculate the product of dissolved oxygen data and cell concentration data to obtain the second interactive feature data. ; Q4: Calculate the product of the miscellaneous bacteria content index data and the foam level data to obtain the third interactive feature data; Q5: Package the first interactive feature data, the second interactive feature data, the third interactive feature data, and the tank pressure data to obtain the feature vector.

3. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, The steps for optimizing the fermentation risk prediction module based on historical batch success and failure data include: W1: Retrieves a set of historical fermentation risk prediction values ​​and historical actual risk values ​​stored in the database; W2: The loss function is obtained by calculating the square of the historical true risk value minus the historical fermentation risk prediction value within the same group; W3: Update the weight factors in step five of the fermentation risk prediction module based on the gradient descent method to obtain new weight factors; W4: Output the new weighting factor to the fermentation risk prediction module and use it for the next calculation of the fermentation risk prediction model.

4. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, The steps for handling different stages of the fermentation cycle based on predicted fermentation risk values ​​include: E1: Based on the changing trends of cell concentration data, the fermentation phase is divided into the lag phase, log phase, stationary phase, and death phase; E2: When the predicted fermentation risk value is greater than or equal to the high fermentation threshold, an adjustment instruction is generated; The adjustment command contains a set of characters representing the adjustment of the fermentation temperature data to the target temperature; E3: When the fermentation stage is logarithmic, if the predicted fermentation risk is less than the first adjustment threshold, an increase in feeding report is generated; if the predicted fermentation risk is greater than the second adjustment threshold, a decrease in feeding report is generated. The replenishment report should include a description of the need to increase the replenishment rate. A reduced replenishment report should include an explanation of the need to reduce the replenishment rate. E4: Packaging adjustment instructions, adding replenishment reports, and reducing replenishment reports, resulting in multi-timescale reports.

5. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, The steps for optimizing multi-objective requirements based on risk prediction values ​​include: R1: Multi-objective requirements include maximizing output, minimizing energy consumption, and minimizing byproduct inhibition; R2: Variables for optimizing multi-objective requirements include feed rate, stirring rate, and fermentation temperature setpoint; R3: Based on maximizing output as the primary objective, an optimization problem is established; R4: Solve the optimization problem based on the genetic algorithm to obtain the optimal feeding rate, optimal stirring rate and optimal fermentation temperature setpoints respectively; R5: The optimal feeding rate, optimal stirring rate, and optimal fermentation temperature settings for packaging are used to obtain a co-optimization report.

6. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, Methods for grading risk predictions and analyzing the results include: Based on the risk threshold interval (Y1, Y2); When the predicted risk value is less than Y1, a low-risk report is generated; when the predicted risk value is greater than or equal to Y1 and less than Y2, a medium-risk report and parameter adjustment instructions are generated; when the predicted risk value is greater than or equal to Y2, a high-risk report and stop instructions are generated. The low-risk report indicates that the predicted risk of fermentation is low, and staff should maintain normal monitoring without intervention. The medium-risk report includes parameter adjustment instructions and explanations predicting the level of fermentation risk. Staff should pay attention to system adjustments and go to the fermentation site to confirm. The high-risk report includes a stop order and a statement indicating a high level of predicted fermentation risk. Staff are requested to immediately enter the site to conduct aseptic sampling after confirming that fermentation has stopped, and to activate the fermenter cleaning mechanism. The parameter adjustment command contains a set of characters representing a reduction in fermentation temperature data to ninety percent; The stop command contains a set of characters that represent stopping the fermenter from running; Package low-risk, medium-risk, and high-risk reports to obtain a risk response decision report.

7. The fermentation monitoring system with multi-parameter adaptive control according to claim 1, characterized in that, Methods for processing multi-timescale reports and risk response decision reports include: Monitor multi-timescale reports and risk response decision reports for any adjustment instructions, parameter adjustment instructions, or stop instructions; When there is an adjustment command, parameter adjustment command, or stop command, the adjustment command, parameter adjustment command, or stop command is converted into a PLC-readable signal through the OPC UA protocol and sent to the intelligent controller. The system dataset includes fermentation dataset, environmental dataset, feature vectors, fermentation risk prediction values, multi-timescale reports, collaborative optimization reports, and risk response decision reports.

8. A fermentation monitoring method with multi-parameter adaptive control, implemented according to any one of claims 1-7, characterized in that, The work includes the following steps: S1: Collect fermentation dataset, which includes pH data, fermentation temperature data, cell concentration data, and ethanol concentration data; S2: Collect environmental datasets, which include bacterial content index data, foam level data, and tank pressure data; S3: Preprocess the fermentation dataset and the environmental dataset, and perform data interaction processing to obtain feature vectors; S4: Process the feature vector to obtain the fermentation risk prediction value; S5: Optimize the fermentation risk prediction module based on historical batch success and failure data to obtain a weight update report; S6: Process the fermentation cycle stages based on the fermentation risk prediction values ​​to obtain multi-timescale reports; S7: Optimize multi-objective requirements based on risk prediction values ​​to obtain a collaborative optimization report; S8: Classify the risk prediction values ​​and analyze the results to obtain a risk response decision report; S9: Stores the system dataset, displays multi-timescale reports, collaborative optimization reports, and risk response decision reports through a visualization panel, and processes the risk response decision reports.

Images