Automatic control system and control method for microbial reaction fermentation

Abstract

The invention relates to the technical field of microbial fermentation control, in particular to an automatic microbial reaction fermentation control system and a control method. The method comprises the following steps: collecting real-time microbial environment parameters and identifying a data multi-axis structure, extracting master control environment interval data, eliminating redundant signals to generate standard microbial environment data, defining a fermentation degree contrast template through fermentation historical records, and constructing a microbial fermentation timeline framework. The method comprises the following steps: acquiring standard environmental data, identifying a fermentation state of the standard environmental data, constructing a simulated microbial environmental field by using the standard microbial environmental data, performing global microbial reaction simulation, determining fermentation reaction structure partitions based on the data, performing difference comparison to generate fermentation difference data, and performing compensation simulation by using the difference data. And automatic fermentation compensation control of microorganisms is realized. According to the invention, the intelligent automatic control and optimization of the microbial fermentation process are realized, the fermentation efficiency and stability are obviously improved, and the manual intervention is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of microbial fermentation control, and in particular to a microbial reaction fermentation automatic control system and a control method. Background Art

[0002] Existing microbial fermentation control methods often rely on experience and manual monitoring, which makes it difficult to obtain and analyze environmental parameters in the fermentation process in real time. Existing technologies have obvious deficiencies in dynamically adjusting the microbial environment and cannot effectively respond to complex changes in fermentation conditions. The monitoring of microbial growth status during fermentation is often not accurate enough, resulting in large fluctuations in product quality and yield, limiting production efficiency and cost control. In addition, the utilization rate of fermentation history records is low, and there is a lack of systematic methods to combine historical data with real-time data, resulting in unscientific judgment and adjustment of fermentation status, and inability to respond to changes within the microbial community in a timely manner. Especially in large-scale industrial production, fermentation differences may have a significant impact on overall production efficiency. Existing technologies fail to effectively distinguish and compensate for these differences, further affecting the stability of the fermentation process. Summary of the Invention

[0003] Based on this, it is necessary to provide a microbial reaction fermentation automatic control system and control method to solve at least one of the above technical problems.

[0004] To achieve the above object, a microbial reaction fermentation automatic control method comprises the following steps:

[0005] Step S1: collecting real-time microbial environmental parameters; identifying the multi-axis structure of the data of the real-time microbial environmental parameters, and extracting the main control environment interval data based on the multi-axis structure of the data; eliminating redundant signals through the main control environment interval data, thereby generating standard microbial environmental data;

[0006] Step S2: defining a fermentation degree control template based on the fermentation history record data; constructing a microbial fermentation timeline framework based on the fermentation degree control template; and identifying the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework;

[0007] Step S3: constructing a simulated microbial environment field using standard microbial environment data; performing a global microbial reaction simulation on the standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field, thereby generating global simulated microbial reaction data;

[0008] Step S4: Identify the microbial community structure in the simulated microbial environment field, and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data; compare the global simulated microbial reaction data for each partition based on the fermentation reaction structure partition, thereby generating fermentation difference data within the microbial tribe;

[0009] Step S5: performing fermentation difference compensation simulation on the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; performing automatic fermentation control according to the simulated compensation data, thereby realizing automatic microbial fermentation compensation.

[0010] The present invention can accurately monitor the key environmental conditions required for microbial growth through the collection of real-time microbial environmental parameters, ensure multi-dimensional structural analysis of data, and thus extract the main control environment interval data. The standard microbial environmental data generated after eliminating redundant signals provides a stable foundation and improves the validity and reliability of the data. The process of defining the fermentation degree control template provides a historical basis for the judgment of the microbial fermentation state, making the control of the fermentation process more scientific and systematic. The constructed microbial fermentation timeline framework effectively integrates historical and real-time data, and enhances the adaptability to dynamic changes in the fermentation process. The construction of the simulated microbial environment field can simulate a more realistic microbial growth environment based on standard microbial environmental data, providing a real basis for subsequent global microbial reaction simulation. The generated global simulated microbial reaction data makes it possible to A deeper understanding of fermentation reaction characteristics can better identify and analyze the characteristics of microbial reactions under different fermentation states. The fermentation reaction structure partitioning determined based on the global simulated microbial reaction data ensures the accurate understanding and control of different reaction areas, improves the difference comparison capability in the reaction process, reveals the fermentation differences within the microbial tribe, and promotes in-depth analysis of the dynamic changes of the microbial community. The differential compensation simulation of the simulated microbial environment field through fermentation difference data can respond to changes in the microbial environment in a timely manner, ensuring that each fermentation area maintains the optimal growth conditions. The generated simulated compensation data provides a scientific basis for the automatic control of the fermentation process, improves the automation level of the fermentation process, reduces the need for manual intervention, ensures the stability and consistency of the fermentation process, and ultimately realizes microbial automatic fermentation compensation and improves the overall fermentation efficiency.

[0011] The present invention also provides a microbial reaction and fermentation automatic control system for executing the microbial reaction and fermentation automatic control method described above. The microbial reaction and fermentation automatic control system comprises:

[0012] The parameter extraction module is used to collect real-time microbial environmental parameters; identify the multi-axis structure of the real-time microbial environmental parameters and extract the main control environment interval data based on the multi-axis structure; eliminate redundant signals from the main control environment interval data to generate standard microbial environmental data;

[0013] A fermentation mapping module is used to define a fermentation degree control template based on fermentation history record data; construct a microbial fermentation timeline framework based on the fermentation degree control template; and identify the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework;

[0014] The simulation field building module is used to build a simulated microbial environment field using standard microbial environment data; based on the microbial fermentation state and the simulated microbial environment field, the global microbial reaction simulation is performed on the standard microbial environment data to generate global simulated microbial reaction data;

[0015] The difference recognition module is used to identify the microbial community structure in the simulated microbial environment field and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data. The global simulated microbial reaction data is compared with each partition according to the fermentation reaction structure partition to generate fermentation difference data within the microbial tribe.

[0016] The intelligent compensation module is used to simulate the fermentation difference compensation of the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; and automatically control the fermentation according to the simulated compensation data, thereby realizing automatic fermentation compensation of microorganisms.

[0017] The present invention realizes the efficient collection and analysis of real-time microbial environmental parameters through the implementation of the parameter extraction module, can accurately identify the multi-axis structure of data and extract the main control environment interval data, and the generated standard microbial environmental data ensures the reliability and effectiveness of subsequent processing. The introduction of the fermentation mapping module enables the effective use of fermentation history record data. The defined fermentation degree control template provides a scientific basis for the judgment of the microbial state. The constructed microbial fermentation timeline framework enhances the understanding and grasp of the dynamic changes of the fermentation process. The simulation field building module builds a simulated microbial environment field and performs global microbial reaction simulation based on standard microbial environment data, realizing a more comprehensive reaction characteristic analysis. The generated global simulated microbial reaction data provides a solid foundation for subsequent difference identification. The difference identification module The structural partitioning of local simulated microbial reaction data ensures the accurate understanding and control of each reaction area. The generated fermentation difference data within the microbial tribe reveals the key differences in the fermentation process. Based on these difference data, the intelligent compensation module can timely perform compensation simulation on the simulated microbial environment field. The generated simulated compensation data provides a scientific basis for the automatic control of the fermentation process, ensuring that each fermentation area is carried out under optimal conditions, and ultimately realizes the automatic compensation and optimization of the microbial fermentation process, improves fermentation efficiency, and reduces the need for manual intervention. The realization of the overall system not only improves the stability and consistency of the microbial fermentation process, reduces production costs, and improves resource utilization, but also promotes the intelligent and automated process of microbial fermentation technology, and meets the modern industry's demand for efficient and precise production. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic flow chart of the steps of an automatic control method for microbial reaction fermentation;

[0019] Figure 2 Detailed implementation flow chart of step S2;

[0020] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0022] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0023] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0024] To achieve this, please refer to Figures 1 to 2 , a microbial reaction fermentation automatic control method, comprising the following steps:

[0025] Step S1: collecting real-time microbial environmental parameters; identifying the multi-axis structure of the data of the real-time microbial environmental parameters, and extracting the main control environment interval data based on the multi-axis structure of the data; eliminating redundant signals through the main control environment interval data, thereby generating standard microbial environmental data;

[0026] Step S2: defining a fermentation degree control template based on the fermentation history record data; constructing a microbial fermentation timeline framework based on the fermentation degree control template; and identifying the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework;

[0027] Step S3: constructing a simulated microbial environment field using standard microbial environment data; performing a global microbial reaction simulation on the standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field, thereby generating global simulated microbial reaction data;

[0028] Step S4: Identify the microbial community structure in the simulated microbial environment field, and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data; compare the global simulated microbial reaction data for each partition based on the fermentation reaction structure partition, thereby generating fermentation difference data within the microbial tribe;

[0029] Step S5: performing fermentation difference compensation simulation on the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; performing automatic fermentation control according to the simulated compensation data, thereby realizing automatic microbial fermentation compensation.

[0030] The present invention can accurately monitor the key environmental conditions required for microbial growth through the collection of real-time microbial environmental parameters, ensure multi-dimensional structural analysis of data, and thus extract the main control environment interval data. The standard microbial environmental data generated after eliminating redundant signals provides a stable foundation and improves the validity and reliability of the data. The process of defining the fermentation degree control template provides a historical basis for the judgment of the microbial fermentation state, making the control of the fermentation process more scientific and systematic. The constructed microbial fermentation timeline framework effectively integrates historical and real-time data, and enhances the adaptability to dynamic changes in the fermentation process. The construction of the simulated microbial environment field can simulate a more realistic microbial growth environment based on standard microbial environmental data, providing a real basis for subsequent global microbial reaction simulation. The generated global simulated microbial reaction data makes it possible to A deeper understanding of fermentation reaction characteristics can better identify and analyze the characteristics of microbial reactions under different fermentation states. The fermentation reaction structure partitioning determined based on the global simulated microbial reaction data ensures the accurate understanding and control of different reaction areas, improves the difference comparison capability in the reaction process, reveals the fermentation differences within the microbial tribe, and promotes in-depth analysis of the dynamic changes of the microbial community. The differential compensation simulation of the simulated microbial environment field through fermentation difference data can respond to changes in the microbial environment in a timely manner, ensuring that each fermentation area maintains the optimal growth conditions. The generated simulated compensation data provides a scientific basis for the automatic control of the fermentation process, improves the automation level of the fermentation process, reduces the need for manual intervention, ensures the stability and consistency of the fermentation process, and ultimately realizes microbial automatic fermentation compensation and improves the overall fermentation efficiency.

[0031] In an embodiment of the present invention, the method for automatically controlling microbial reaction and fermentation comprises the following steps:

[0032] Step S1: collecting real-time microbial environmental parameters; identifying the multi-axis structure of the data of the real-time microbial environmental parameters, and extracting the main control environment interval data based on the multi-axis structure of the data; eliminating redundant signals through the main control environment interval data, thereby generating standard microbial environmental data;

[0033] In this embodiment, data collection is completed by a multi-point online monitoring sensor array installed inside the bioreactor. The sensor array consists of a dissolved oxygen sensor, a pH sensor, a temperature probe, an infrared carbon dioxide concentration monitor, a conductivity monitoring probe, and a biomass concentration laser scanner. Each sensor communicates with the PLC control system through the RS485 industrial bus. The sensor signal enters the data acquisition module at a frequency of 10 times per second to form a real-time microbial environmental parameter data stream, which then enters the structure recognition module based on the multi-axis feature extraction algorithm. The six-dimensional monitoring data is integrated into the data stream through the main control axis judgment rule. First, it is normalized into a three-axis collaborative structure data stream with dissolved oxygen, pH value and temperature as the main axes. The dimensional signals of other parameters with a fluctuation rate of less than 0.05 in a continuous 50-frame sampling period are screened out, marked as redundant axes and eliminated, and finally the data are weighted averaged based on the three axes. The maximum gradient change rate and dynamic interval stability value of the sliding window of each axis within 20 seconds are extracted to form the main control environment interval data. The control limit-based elimination model is used to remove data abnormal jump points and low-credibility segments in the main control interval. Finally, the remaining data is smoothed by Kalman filtering to output the standard microbial environment data stream.

[0034] Step S2: defining a fermentation degree control template based on the fermentation history record data; constructing a microbial fermentation timeline framework based on the fermentation degree control template; and identifying the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework;

[0035] In this example, a fermentation history database was constructed by extracting data samples from 180 days of similar fermentation batch experimental data. The database uses each fermentation batch as a data unit and includes the fermentation start time, fermentation end time, environmental parameter change trajectory, and terminal metabolite concentration detection results. Batches with a final metabolite concentration greater than 98% purity are considered valid samples. A fermentation stage discrimination algorithm based on clustering and segment matching is used to divide the time series data of the valid samples into similar segments. The environmental parameter change patterns within each time period are classified into five fermentation degree types using the K-means clustering method. Each type includes upper and lower limit constraints on the characteristic indicator range. A fermentation degree control template is constructed. This template is used to perform interval judgment on the three main control indicators of pH, dissolved oxygen, and temperature in the current standard microbial environmental data to identify the fermentation stage corresponding to the current data in the time dimension. The dynamic programming time series matching algorithm (DTW) is used to perform optimal path comparison between the standard data and the template data. The position of the standard data in the five stages is mapped based on the comparison path position, and the microbial fermentation state to which the standard microbial environmental data belongs is ultimately determined to be the "acidification-enhanced mid-stage reaction stage."

[0036] Step S3: constructing a simulated microbial environment field using standard microbial environment data; performing a global microbial reaction simulation on the standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field, thereby generating global simulated microbial reaction data;

[0037] In this embodiment, the obtained standard microbial environment data and the aforementioned determined microbial fermentation state are used to call a high-performance computing platform to construct a simulated microbial environment field. The simulation process uses a three-dimensional spatial reaction grid structure, and each grid node updates the reaction state once every 10 milliseconds. The CFD (Computational Fluid Dynamics) model is used to model the internal fluid state of the fermentation tank. At the same time, the microbial metabolic reaction dynamic parameters, including glucose consumption rate, lactate production rate, protein expression starting concentration, and rate variation coefficient, are loaded as basic parameters. Based on the initial concentration range of the acidification-enhanced mid-stage reaction, the standard microbial environment data are mapped to the simulation input boundary conditions. The microbial reaction simulation process is run using an algorithm based on the reaction-diffusion-migration model. The model is advanced forward along the simulation timeline to simulate the diffusion and reaction results of the fermentation substances in the simulated microbial environment field. Finally, a global simulated microbial reaction data matrix is ​​generated within a 120-second simulation cycle. The data matrix is ​​organized by spatial position index in units of 0.5 mm cubic grids, and each grid contains three sets of data: reaction rate, product concentration, and pH change rate.

[0038] Step S4: Identify the microbial community structure in the simulated microbial environment field, and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data; compare the global simulated microbial reaction data for each partition based on the fermentation reaction structure partition, thereby generating fermentation difference data within the microbial tribe;

[0039] In this embodiment, based on the generated global simulated microbial reaction data matrix, the reaction structure partitioning algorithm module is called to perform spatial partitioning processing. The module first performs a three-dimensional difference calculation on the reaction rate gradients of all spatial grid points, identifies continuum unit blocks with local reaction rate increments greater than 5%, and uses them as candidate high reaction areas. At the same time, areas with rates less than 20% of the average are screened out as weak reaction areas. A reaction structure partition map is generated based on the boundary information of the two areas. The partition index positioning function is used to extract the microbial reaction data within each partition. The average pH change rate and metabolite concentration in each partition are compared. The Z-score normalization method is used to evaluate the degree of difference. All interval data with differences significantly greater than the standard deviation are extracted and organized into fermentation difference data within the microbial tribe. Each data unit is accompanied by spatial position coordinates, fermentation parameter name, timestamp, and parameter change degree value, ultimately forming a structured JSON format data packet for subsequent compensation simulation module calls.

[0040] Step S5: performing fermentation difference compensation simulation on the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; performing automatic fermentation control according to the simulated compensation data, thereby realizing automatic microbial fermentation compensation.

[0041] In this embodiment, the generated fermentation difference data is input into the fermentation difference compensation simulation engine, and the scene reconstruction module is used to reconstruct the local high fermentation response area based on the original partition grid in the simulated microbial environment field, and the perturbation factor parameters are injected, such as the local temperature is increased by 2 degrees Celsius, the dissolved oxygen is increased by 5 mg per liter, and the glucose feed rate is increased by 10%. The perturbation simulation module is used to independently perform a 120-second simulation on each group of factors and output dynamic reaction data. The factor parameter combination with a response value significantly close to the target metabolite concentration threshold is then screened and used as the compensation control input to generate simulated compensation data. The data includes the reaction zone position coordinates, compensation factor type, value, action timing and corresponding predicted reaction results. Finally, the simulated compensation data is sent to the fermentation control system, and the on-off rhythm of the precision heater, oxygen pump, and nutrient pump is adjusted through the PLC module to adjust the local microbial environment state parameters so that the physical fermentation process is close to the optimal simulation feedback path in real time, thereby realizing a fully closed-loop drive response mechanism for the microbial automatic fermentation compensation control process.

[0042] Preferably, step S1 includes the following steps:

[0043] Step S11: integrating the original data of temperature, pH value, dissolved oxygen, and pressure collected by the sensor into real-time microbial environmental parameters;

[0044] Step S12: performing signal mean level conversion on the real-time microbial environmental parameters to obtain a multi-axis conversion matrix, wherein the conversion step range is 0.2-0.5 units;

[0045] Step S13: performing difference convergence filtering on the multi-axis conversion matrix to generate the main control section stability parameter, wherein the filter window width range is 5-15 groups of data, and the filter threshold is that the standard deviation of each axis does not exceed 0.35 units;

[0046] Step S14: extracting the main control environment interval data from the main control section stability parameters, wherein the time span of the main control section is not less than 20 minutes;

[0047] Step S15: determining the mutation environment parameters in the real-time microbial environment parameters based on the main control environment interval data;

[0048] Step S16: Eliminate redundant signals in the real-time microbial environmental parameters based on the mutant environmental parameters, thereby generating standard microbial environmental data.

[0049] In this embodiment, in the process of converting the original values ​​of temperature, pH value, dissolved oxygen concentration and pressure in the tank collected by the sensor into real-time microbial environmental parameters, four types of independent Internet of Things high-precision digital sensors are used, namely temperature sensor model PT100, measurement accuracy of ±0.1 degrees Celsius, pH sensor model PH201-industrial type, measurement accuracy of ±0.01, dissolved oxygen sensor model DO-850, and pressure sensor model DPS-400. The data acquisition cycle is uniformly set to once per second. Each type of sensor is connected to the RS485 serial port, and is connected to the edge computing module after being encapsulated by the Modbus protocol. The format of the collected original numerical data is unified as a data sequence of timestamp plus 4 dimensions, which constitutes a real-time microbial environmental parameter original data vector sequence. The processing capability of the DSP chip embedded in the edge module is called to perform continuous sliding mean processing on the real-time data sampled per second, and the original data within every 60 seconds is locally segmented. The data were normalized and linearly differentiated using the mean of the past 60 seconds as the reference baseline. The conversion step was 0.3 units, corresponding to 0.3 degrees Celsius in the temperature dimension, 0.3 pH units in the pH dimension, 0.3 mg / L in the dissolved oxygen dimension, and 0.3 bar in the pressure dimension. After processing, a multi-axis transformation matrix consisting of four-dimensional data was generated. Each row corresponds to the four-dimensional mapping value at a time point. The matrix structure dimension is T×4, where T is the number of data in the processing time window. Each data item has been stripped of its absolute value information and converted into a relative mean deviation vector. A difference convergence filtering operation was performed using an improved median filter. A sliding window was constructed in each axial dimension. The window width was fixed to 10 groups of data. The standard deviation of the dimension was calculated in each group. If the standard deviation of any group of 10 consecutive groups of data exceeded 0.35 units, the window was discarded and did not participate in the next round of screening. The standard deviation of each axis did not exceed 0.Under the condition of 35 units, the current window data is retained as the stable data segment input to the main control segment parameter pool, and the mean operation on all dimensions is performed on each window that passes the screening, and the result is recorded as the stable parameter. This process continues until the entire data segment is traversed. The stable parameter quadruple output in each round constitutes a stable parameter sequence, which is stored as the main control segment stable parameter data set. The main control environment interval data extraction operation is executed, and the sliding sampling time window length is set to 30 minutes. The stable parameter data is extracted 6 times per minute to form 180 groups of stable parameter vectors. The window sliding algorithm based on time sequence is used to determine whether it is within the continuous time segment. The variance of all stable parameters is lower than the aforementioned threshold and the values ​​have no sudden fluctuations. After confirming that the conditions are met, the stable parameter data in the continuous time period is extracted as the main control environment interval data and the corresponding start and end time is recorded. If there is a mutation point in any sampling period within 30 consecutive minutes, the entire interval will not be adopted. The extraction operation takes the longest stable time as the priority extraction strategy, and finally obtains multiple main control environment segments. The time span of each segment is not less than 20 minutes. Each group of data is numbered in timestamp order. After extracting the main control environment interval data, enter the mutation environment parameter identification stage, and use the real-time microbial environmental parameters of the previous and next time periods and the main control environment interval to identify the mutation environment parameters. The vector comparison of the data between time points is performed, and the distance between each time point and the mean vector of the main control interval is calculated using the Euclidean distance method. The distance threshold is set to 3 times the standard deviation of the main control mean deviation. When the Euclidean distance between the real-time parameter and the main control mean exceeds the threshold, it is determined to be a mutation point. In the 5-minute period before and after the mutation point, continue to track whether there is a continuous trend of such deviation expansion. If there is a continuous abnormal trend, it is identified as a mutation environment parameter interval. All abnormal dimension parameters are recorded as mutation environment parameters, and mutation type labels are generated, including three categories: increase, decrease, and unstable fluctuation. After obtaining the mutation environment parameters, the corresponding mutation dimensions in the real-time microbial environmental parameters are categorized. All data segments were processed to remove redundant signals. Within the time period identified by the mutation parameters, a regression filtering algorithm was combined with the wavelet transform boundary reconstruction method to perform local reconstruction of the dimension containing the mutation signal, separating and removing high-frequency perturbations from low-frequency fluctuations. After restoring the original signal trend line, the remaining dimensional data was combined to construct a new four-dimensional parameter vector. This vector reconstructed a stable data structure after redundancy removal, namely the standard microbial environmental data. The standard microbial environmental data format remains consistent with the original raw data, including timestamps and four-dimensional numerical information. Each data item also carries a quality label to indicate whether it has undergone mutation denoising.

[0050] Preferably, step S2 includes the following steps:

[0051] Step S21: querying the fermentation history data of the microorganism through the log; mapping the fermentation process characteristic intervals according to the fermentation history data to define a fermentation degree control template, wherein the fermentation history data has a temperature range of 28-38°C, a pH range of 5.5-7.2, a dissolved oxygen range of 15-85%, and a pressure range of 0.8-1.2 bar;

[0052] Step S22: Synchronously pairing the fermentation degree control template with the stage anchor points to obtain timeline pairing anchor points; constructing a microbial fermentation timeline framework through the timeline pairing anchor points, wherein the time interval corresponding to each stage in the microbial fermentation timeline framework ranges from 8 to 30 hours, and the state anchor point matching deviation is controlled within a range of 2 hours;

[0053] Step S23: projecting the fermentation characteristic data in the standard microbial environment data into the microbial fermentation time frame, and performing fermentation characteristic pairing to obtain paired fermentation characteristics;

[0054] Step S24: determining the corresponding microbial fermentation state in the microbial fermentation timeline frame based on the paired fermentation characteristics.

[0055] In this example, a high-speed data interface was established between the historical data query module deployed at the fermentation tank control terminal and the microbial fermentation data management platform. The strain number was set to XM2301 using the interface parameter setting function, and the fermentation batch time range was limited to the past 90 days. A total of 18 batches of historical fermentation record data related to the XM2301 strain were extracted. Each batch of data included the temperature, pH value, dissolved oxygen concentration, and tank pressure values ​​at consecutive sampling time points. The sampling interval was set to 5 minutes. A four-dimensional time series matrix was constructed for each batch. Each data item was traversed to filter out records with a temperature range of 28 to 38 degrees Celsius, a pH range of 5.5 to 7.2, a dissolved oxygen concentration range of 15% to 85%, and a tank pressure range of 0.8 to 1.2bar, after eliminating the fragmented data that does not meet the parameter restrictions, the data within each hour are merged and averaged as the characteristic time point, and the intervals of the fermentation yield growth inflection points in each batch of data are counted, the start time and end time of these intervals are recorded, and the historical fermentation process is divided into five stages: inoculation preparation period, adaptation growth period, efficient growth period, stable period and end period, combined with the trend of curve slope change. The interval mean of each stage is used as the characteristic parameter of the stage, and a fermentation degree control template is constructed. The output is a structured data format table with fields including stage name, interval mean temperature, interval mean pH, interval mean dissolved oxygen and interval mean tank pressure. The constructed fermentation degree control template is imported into the anchor point pairing system, and by setting The sliding step of the time window is set to 1 hour, and the starting and ending points of the five stages in the historical data of each fermentation batch are compared with the data intervals of 2 hours forward and backward. The time period with a parameter change of more than 10% per hour is marked as a candidate stage anchor point, and the candidate anchor points of the same stage in multiple batches are clustered. The number of cluster centers is set to 5 using the K-means algorithm, and the central time with the highest data overlap in each category is extracted as the final anchor point position. All anchor point control time values ​​are paired to a standard time axis using the linear time stretching method. The total length of the standard time axis is divided into 150 hours, with a time interval of 30 hours. The time of all anchor points is mapped to the standard time axis to construct the microbial fermentation time. The timeline framework is used, in which each stage corresponds to a time period, and the anchor point deviation tolerance range is set to plus or minus 2 hours. If the anchor point deviation exceeds this range, no pairing operation will be performed on it. The standard microbial environment data obtained from the previous processing steps are sequentially imported into the fermentation feature recognition module, and the four dimensions of temperature, pH, dissolved oxygen, and tank pressure in the hourly sampling values ​​are extracted as fermentation feature data fields. These data are mapped and matched with the standardized microbial fermentation timeline framework according to their sampling timestamps. The window alignment method is used to match each group of sampling time periods to the corresponding standard stage intervals in the fermentation timeline. The average value vector in each time period is taken and the similarity is calculated with the corresponding stage template value in the timeline framework. The similarity calculation adopts the cosine similarity function The matching threshold is set to 0.85. Similarities above this value are considered to be successfully paired fermentation features. Successfully matched environmental parameter groups are labeled as paired fermentation features, and their corresponding stage names, sampling time period numbers, and offset values ​​are recorded. Paired fermentation features are then imported into the microbial fermentation state judgment module. The judgment rule is set as follows: if the current feature parameter group successfully matches any standard fermentation stage template and the offset is less than the set tolerance, the corresponding fermentation state is directly marked based on the stage name of the matched template. If the current feature parameter matches the same stage template in three consecutive time windows and the offset is less than 1 hour, the current time point is determined to be a stable fermentation stage. If the matching results show a cross-stage distribution and no feature has a value higher than 0.If the similarity is 85, it is recorded as a state transition zone. If there is no successful match result in three consecutive time windows and the environmental parameter fluctuation range exceeds the set fluctuation threshold, such as the temperature fluctuation exceeds 3 degrees Celsius or the pH fluctuation exceeds 0.4 units, it is judged as an unstable abnormal state. All fermentation status results are stored in the state tracking database, with fields including current time, matching stage, state type, offset value, matching confidence and abnormal flag field.

[0056] Preferably, in step S3, building a simulated microbial environment field using standard microbial environment data includes:

[0057] Extract various environmental characteristic parameters from standard microbial environmental data;

[0058] Reconstruct the spatial grid environment from standard microbial environmental data according to environmental characteristic parameters;

[0059] Identify local heterogeneous features in a spatial grid environment;

[0060] Correcting local heterogeneous features based on environmental feature parameters to generate corrected environmental features;

[0061] Construct a simulated microbial environment field by correcting environmental characteristics.

[0062] In this example, environmental characteristic parameters such as temperature, pH value, dissolved oxygen concentration, carbon dioxide concentration, pressure, and stirring rate were extracted from standard microbial environmental data. A high-precision sensor array was used for data acquisition, and the sampling frequency was set to once per second. The raw data was filtered to eliminate noise interference, and each environmental parameter was mapped to a standardized range of 0 to 1 using a normalization algorithm. Principal component analysis (PCA) was used to analyze the data. The dimensionality reduction of multidimensional parameter data was performed by PCA (Procedure Analysis, PCA), key eigenvectors were extracted, and a set of characteristic parameters was formed. The three-dimensional spatial model inside the fermentation tank was defined, and the size was set to 2 meters in length, 1.5 meters in width, and 1.5 meters in height. The space was divided into uniform cubic grids with a side length of 0.1 meters, forming a 20×15×15 spatial grid structure. The environmental characteristic parameters corresponding to each sampling time point were distributed between the spatial grid points through weighted cubic spline interpolation. The weights in the interpolation algorithm were dynamically adjusted according to the sensor distance and stirring rate to ensure the continuity of spatial data. After the grid environmental parameter matrix was constructed, the multi-threaded computing framework was used to realize data parallel processing. The local heterogeneous feature recognition algorithm based on neighborhood statistical analysis was applied to the constructed spatial grid environmental data. The neighborhood of each grid point was defined as 27 neighboring grid points with a side length of 0.3 meters. The local coefficient of variation of the environmental parameters in each neighborhood was calculated. Variation (LCV), heterogeneous features are defined as grid points with LCV exceeding 0.15. During the identification process, the three key parameters of temperature, pH and dissolved oxygen are evaluated independently. The heterogeneous grid points are aggregated into heterogeneous blocks by combining the spatial clustering algorithm, and the block volume, center position and heterogeneous intensity index are calculated. The heterogeneous intensity is defined as the ratio of the difference between the maximum and minimum values ​​of the parameters in the block to the mean. Based on the identified local heterogeneous features, the environmental feature correction is performed using the multidimensional weighted Gaussian filtering algorithm. The Gaussian kernel function with a kernel width of 0.15 meters is selected to calculate the weighted average of the grid points in each heterogeneous block. The weight is a combination of the distance weight and the inverse weight of the heterogeneous intensity. During the correction process, spatial smoothing is performed independently on parameters such as temperature, pH and dissolved oxygen. The correction data replaces the original spatial grid point parameters to ensure that the parameters are continuous and smoothly distributed in space. The correction error is controlled within ±0.005. The multi-threaded asynchronous computing mode is used to improve the computing efficiency. The corrected three-dimensional environmental feature data is imported into the simulated microbial environmental field modeling module. The finite element analysis (Finite element method) based on Finite element method is used to calculate the weighted average of the grid points in each heterogeneous block. Element Analysis (FEA) method is used to construct a three-dimensional simulation grid containing tetrahedral elements. The element size is set to 0 according to the spatial grid spacing.1 meter, environmental parameters are input as node attributes to achieve continuous parameter interpolation simulation, boundary conditions are set as fixed values ​​of fermentation tank wall temperature and pressure, fluid dynamics parameters are input as dynamic boundaries, and a coupled microbial dynamics model and environmental field model are used to calculate the spatiotemporal evolution of environmental parameters during microbial fermentation through a numerical solver, outputting three-dimensional continuous distribution of simulated environmental field data such as temperature, pH, and dissolved oxygen.

[0063] Preferably, in step S3, performing global microbial reaction simulation on standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field includes:

[0064] Projecting the microbial fermentation state onto the simulated microbial environment field to obtain the fusion fermentation environment state;

[0065] Mapping the fusion fermentation environment state to the reaction tendency field to obtain reaction tendency field data;

[0066] Retrieve local microbial fermentation diffusion paths by comparing reaction tendency field data with standard microbial environment data;

[0067] Build collaborative feedback links based on local microbial fermentation and diffusion pathways;

[0068] The spatiotemporal evolution driving is performed according to the collaborative feedback link, thereby obtaining the spatiotemporal evolution driving data;

[0069] The local microbial responses are simulated by driving data based on spatiotemporal evolution, and then superimposed and integrated into global simulated microbial response data.

[0070] In this embodiment, the microbial fermentation state data includes multidimensional dynamic indicators such as microbial population density, metabolic rate, product concentration and nutrient consumption rate. The simulated microbial environment field is composed of a three-dimensional spatial grid environment constructed by the aforementioned corrected environmental characteristics. The fermentation state data is mapped to the corresponding spatial grid position using a vector projection method. According to the spatial coordinate matching principle, each microbial state vector is combined with the spatial grid environmental characteristics to form a fused fermentation environment state matrix. The nearest neighbor interpolation method is used in the projection process to ensure data alignment. The data fusion result is stored in a tensor structure to facilitate subsequent complex calculations. The data dimensions include spatial coordinates and multidimensional state characteristics. Based on the fused fermentation environment state, a nonlinear mapping function is used to construct a reaction tendency field. The fused environment state matrix is ​​mapped through a multi-layer perceptron (MLP, Multi-LayerPerceptron) neural network. The neural network input is the multidimensional feature vector of the fused state. The network structure includes three hidden layers, with the number of hidden layer nodes being 128, 64, and 32, respectively. The activation function uses ReLU (Rectified Linear Unit). The output layer corresponds to the tendency value at each spatial grid point of the reaction tendency field. The tendency value reflects the reaction activity of the microorganism at a specific spatial location. The dataset used to train the neural network is derived from the correspondence between environmental parameters and reaction products collected from historical fermentation experiments. The training process uses the Adam optimization algorithm, and the learning rate is set to 0.001, the number of training iterations is 5000, and the reaction tendency field data matrix is ​​finally obtained. Based on the reaction tendency field data, a graph theory method is used to establish a connection network of spatial grid points. Each grid point is used as a network node, and the tendency value is used as the node weight. The breadth-first search (BFS) algorithm is used to identify the microbial fermentation diffusion path starting from the high reaction tendency node. The diffusion path is defined as a continuous node sequence with increasing tendency values ​​in the adjacent nodes. The path length is limited to no more than 30 nodes. During the path search process, spatial distance constraints are combined to ensure the physical continuity of the path. The identified local microbial fermentation diffusion path set is stored as a directed graph structure file. The path node sequence records the spatial coordinates and corresponding reaction tendency values ​​on the path in detail. The path search algorithm is accelerated using parallel computing technology. Based on the identified diffusion path, a collaborative feedback link is constructed. The spatial nodes in the path are connected in sequence to form a link unit. The link unit parameters include the reaction intensity difference between nodes, the distance interval and the time delay estimation. The time delay is obtained through regression analysis based on the historical fermentation process data. The link construction adopts the dynamic time warping algorithm (DTW) Warping is used to match path node time series to ensure temporal continuity and synchronization within the link. The link structure is stored in a multidimensional array, recording spatial node identifiers, inter-node transmission strength, and overall link feedback efficiency. Collaborative feedback links are managed through a graph database, supporting cross-path analysis and dynamic adjustment of feedback strength. Based on the collaborative feedback link data, a spatiotemporal evolution-driven model is established. Partial differential equations (PDEs) are used to describe the temporal and spatial variations in the concentration of microbial reaction substances. The equation parameters are calculated from the transmission strength and time delay within the link. The model is discretized using the finite difference method (FDM) in the spatial and temporal domains, with a time step of 0.01 hour and a spatial step of 0.1 meter, using an implicit iterative method to solve the evolution equations and obtain a spatiotemporal evolution driving data matrix. The matrix dimensions correspond to spatial grid points and time series, simulating the diffusion and feedback mechanisms of microbial reactions. Numerical calculations are performed in parallel using high-performance computing clusters to ensure accuracy and efficiency. The driving data is stored as a multidimensional tensor and input into the local microbial reaction simulation module. The microbial growth rate, metabolite production rate, and substrate consumption rate at each spatial grid point are calculated. The reaction model is based on the Monod equation and introduces environmental factor correction terms. The simulation results of each local grid point are superimposed and integrated to form global microbial reaction data. The integration process uses a weighted accumulation method, with weights adjusted based on the spatial grid point volume and reaction activity. The integration results are output as three-dimensional time series data in a format compatible with the data interface of the automatic control system for subsequent fermentation parameter adjustment and process optimization. The calculation process adopts a pipelined parallel execution.

[0071] Of particular importance is the mapping of the fusion fermentation environmental states to the reaction propensity field, including:

[0072] Perform feature decomposition on the fusion fermentation environment state to obtain local fermentation characteristics;

[0073] Perform feature correlation measurement on local fermentation features to generate feature co-distribution data;

[0074] Performing structure remapping based on feature co-distribution data to obtain response matching data;

[0075] The behavior block indexing of the fusion fermentation environment state is performed according to the response matching data, thereby generating fermentation block projection data;

[0076] Based on the fermentation block projection data, block density normalization was performed to obtain the reaction tendency field data.

[0077] In this embodiment, the principal component analysis (PCA) method is used to perform feature decomposition on the fusion fermentation environment state, and the eigenvectors and eigenvalues ​​of the data are calculated using the covariance matrix. The main local fermentation features are extracted by selecting the top N principal components with a cumulative contribution rate of more than 90%. These features represent the most significant variation patterns in the data. The decomposition process is achieved by singular value decomposition (SVD), which ensures that the eigenvectors after dimensionality reduction retain key dynamic information of the environment and fermentation state while reducing data redundancy. The feature decomposition results are saved in the form of a feature matrix with a dimension of the number of spatial nodes multiplied by the number of principal components. For the extracted local fermentation features, the correlation measure between the features is calculated, and the mutual information (Mutual Information) indicator is used to quantify the statistical dependence between each feature pair. The calculation process is based on the discretized feature data and uses kernel density estimation (KDE,The Kernel Density Estimation method is used to estimate the joint probability distribution and marginal probability distribution, and then the mutual information value is obtained to generate a feature co-distribution data matrix. Each element in the matrix represents the correlation strength of the corresponding feature pair. The correlation measurement result is normalized to the range of 0 to 1 through regularization to ensure that the feature co-distribution data reflects the collaborative relationship between features. Based on the feature co-distribution data, the graph embedding technology is used to achieve structural remapping and construct a feature collaborative network. The nodes are local fermentation features and the edge weights are feature associations. The feature network is partitioned using the spectral clustering algorithm. The partitioning process is achieved by calculating the eigenvectors of the Laplace matrix. The partitioning results are used to adjust the spatial mapping relationship between features and complete the remapping of the original feature space. The response matching data is output. The data format is the coordinates of the remapped features in the low-dimensional space and their mutual relationship parameters. The mapping process is based on an iterative optimization algorithm with the goal of minimizing the mapping error and maintaining the consistency of the distance between features. After the structural remapping is completed, the response matching data is stored in a multidimensional array. Based on the response matching data, the behavior block indexing operation is performed and the density peak clustering is used. Clustering) algorithm identifies high-density feature areas in the fermentation process. In the clustering process, the local density of each feature point is first calculated. The density calculation is based on the Gaussian kernel function, and the kernel width parameter is set to 5% of the average distance in the data space. Then, the center point with a farther distance and higher density is determined as the cluster center. Each feature point is assigned in the direction of decreasing density according to the center point to generate multiple fermentation blocks. The features in the block show highly correlated behavior patterns. The indexing results are stored in a block label matrix, corresponding to each spatial grid point and its features. The block indexing data is the fermentation block projection data, which is used to describe the spatial distribution and dynamic characteristics of different functional blocks in the fermentation environment. Based on the fermentation block Projected data is then subjected to density normalization for each block. Using the maximum-minimum normalization method, the block density values ​​are linearly mapped to the range of 0 to 1. The specific calculation formula is: the normalized value equals (current block density minus minimum density) divided by (maximum density minus minimum density). During the normalization process, the block density at each time node is processed independently to ensure the continuity and consistency of the time series data. The normalized data serves as the basic data for the reaction tendency field. The reaction tendency field data is stored in the form of a three-dimensional tensor, with dimensions including spatial coordinates, time nodes, and normalized density values. The storage format is compatible with the automatic control system interface to ensure subsequent dynamic monitoring and adjustment of microbial reaction tendencies.

[0078] Preferably, in step S4, identifying the microbial community structure in the simulated microbial environment field and determining the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data includes:

[0079] Hierarchical identification of microbial population activity trajectories in simulated microbial environment fields;

[0080] Perform temporal clustering of microbial population activity trajectories to generate multiple microbial subpopulation clustering areas;

[0081] The community structure was analyzed by microbial subpopulation aggregation areas to determine the microbial community structure;

[0082] The global simulated microbial reaction data is divided into multi-dimensional grids to obtain the microecological activity heat, metabolite accumulation rate and growth curve transfer rate;

[0083] A three-axis indicator mapping matrix was constructed based on the microecological activity heat, metabolite accumulation rate, and growth curve transfer rate;

[0084] The sensitivity of different microbial subpopulation clusters was quantified based on the three-axis index mapping matrix, and the dynamic fermentation response diagram of each local microbial unit was drawn;

[0085] The microbial community structure was spatially overlaid with the dynamic fermentation response map, and high response trajectory areas were identified;

[0086] The standard microbial environment data are divided into multiple fermentation reaction structure partitions based on the high response trajectory areas.

[0087] In this embodiment, a high-dimensional time series analysis module is deployed to read the simulation operation log for each spatial grid point in the simulated microbial environment field, extract dynamic data such as microbial community density, metabolic rate, and energy release, and establish a microbial activity trajectory sequence for each grid point along the time axis. The trajectory data is recorded with a time stamp of 10 minutes. The trajectory length covers the entire simulation process. Subsequent trajectories are cross-linked with position coordinates to form the spatiotemporal distribution of microbial activity in the entire field. The structure is stored as a three-dimensional tensor and imported into the spectral clustering preprocessing system. The above trajectory tensor is subjected to time series clustering analysis. First, a similarity matrix is ​​constructed based on the spectral clustering method. The similarity is based on the DTW (Dynamic Time Warping) between the trajectory curves. Warping) distance and trajectory starting density difference are calculated together, and the weighted similarity matrix is ​​formed by combining the adjacency relationship of the spatial position of each trajectory. Then the eigenvalue of the Laplace matrix is ​​calculated and the number of clusters is determined by using the maximum eigenvalue gap. Finally, the trajectory data is divided into multiple microbial subpopulation aggregation areas. Each area corresponds to a cluster of trajectories and is automatically labeled. The clustering results are stored in the spatial coordinate table and the trajectory group table. The community structure analysis operation is performed on the above trajectory clustering results. The boundary coordinates and group activity indicators of all member trajectories of each subpopulation aggregation area are extracted, and then the clusters are divided according to the group density. The shape of the community boundary was determined by fitting a two-dimensional Gaussian model with a curve. The community center, average density, and edge density gradient were calculated to construct a microbial community structure model representing each subpopulation. The structure model included the community number, spatial coordinate range, trajectory density, and metabolic representative parameters. The structure model was saved as a JSON structure, providing a basic community framework for subsequent partitioning. When multidimensional gridding was performed on the global simulation reaction data, the environmental space was divided into community units and out-of-community units according to the spatial boundaries in the community structure model. Each unit was further subdivided into fixed-size subgrids with a side length of 0.The 5 mm microgrid is used to calculate the microecological activity heat (i.e., carbon source consumption rate and enzymatic activity integral value), metabolite accumulation rate (accumulation of first-order difference values ​​of multiple metabolite time series concentrations), and growth curve transfer rate (rate of change of bacterial density ratio) in each microgrid. The three indicators are generated into three-dimensional spatial matrices and fused with the corresponding community structure coordinates. When constructing a three-axis mapping system based on the above indicator matrix, the three indicators are mapped to a three-channel tensor structure according to a unified spatiotemporal index. The tensor size corresponds to the grid layout of the entire simulation field. The indicator values ​​are converted to the range of 0 to 1 through normalization function processing. The channel weights are then approved and tensors are fused to generate a multidimensional indicator mapping matrix. Each element in the matrix includes a combination of three-axis values. It is stored in a GPU-accelerated model for high-concurrency access and the main synergistic trends of the mapping can be verified by principal component analysis. Based on the mapping results and community structure labels, the differential sensitivity of different subpopulation aggregation areas is quantified. First, all grid point mapping values ​​are extracted by region and a curve time series is generated. The change amplitude, frequency, and rate of each time series are calculated. Then, variance analysis statistical analysis is performed within the region. Regional sensitivity parameters are generated, and sensitivity scores are aggregated according to preset weights to form regional indicator scores. A dynamic fermentation response map is created by plotting regional indicator curves over time and overlaying regional boundary annotations. The generated dynamic response map is spatially overlaid with the microbial community structure. The response map is mapped back to the original spatial grid according to the spatial labels of the trajectory clusters. Spatial regions with absolute response values ​​1.5 times higher than the population mean are identified as high-response trajectory regions. Connectivity analysis is performed within the boundary intervals to extract the shapes of each response hotspot. Intersecting regions are jointly analyzed to form a composite response structure matrix. High-response regions are labeled based on the response level and their spatial coordinate ranges are extracted. Based on the overlaid response structure matrix, the standard microbial environmental data are spatially partitioned into high-response, medium-response, and low-response regions. The environmental parameters and reaction data are merged to generate a partition label matrix. This partition label is attached to each spatial grid point in the standard environmental data. It serves as a recognizable structural partition unit in fermentation process control, providing fine-grained control region boundaries for subsequent compensation control and parameter adjustment, thus achieving physical spatial classification of the fermentation reaction structure and completing the partitioning process.

[0088] Preferably, in step S4, performing partition-by-partition comparison of the global simulated microbial reaction data according to the fermentation reaction structure partitions includes:

[0089] Perform sequence cross-comparison on the microbial response time series within the fermentation reaction structure partition to obtain partition response cross-data;

[0090] Extracting fermentation difference signal units that exhibit any of the following phenomena: metabolic response delay, product conversion deviation, and growth curve drift from the partition response cross-data;

[0091] Based on the fermentation difference signal unit, a differential representation vector set within the microbial tribe was constructed;

[0092] The differential characterization vector set is differentially labeled according to the response intensity, starting offset point and stable time period, thereby generating fermentation difference data within the microbial tribe.

[0093] In this embodiment, for the microbial response time series within the fermentation reaction structure partition, the time series data corresponding to each partition is first extracted based on the spatial partition label. The time series data covers multi-dimensional indicators such as the metabolic activity of the microbial community, product concentration, and bacterial density. The dynamic time warping algorithm (DTW) is used to calculate the time series data. The time series of different sampling points in the same partition were cross-compared by DTW (Dynamic Warping) to calculate the similarity and offset between the time series. The maximum allowed time offset of the DTW parameter was set to 30 minutes, and the distance metric was Euclidean distance. The partition response cross-data matrix was obtained. Each element in the matrix represented the matching distance of the time series of two sampling points and the corresponding time offset information. The fermentation difference signal units showing any of the phenomena of metabolic response delay, product conversion offset and growth curve drift were extracted from the partition response cross-data. The specific operation was to first set a threshold to screen the units with a time series matching distance greater than 0.2 and a time offset greater than 10 minutes. For metabolic response delay, the peak time difference of metabolic indicator response was calculated to screen time series pairs with a peak time offset of more than 15 minutes. For product conversion offset, the morphological characteristics of the metabolite concentration curves, including the relative changes in peak height, half-peak width and peak area, were compared to screen the units with a relative difference in morphological parameters greater than 20%. For growth curve drift, the exponential growth model of the bacterial density curve was fitted to calculate the fitting parameter difference, and the units with a parameter change of more than 0.1 were selected. Units with one condition were labeled as fermentation differential signal units. The signal unit location information and time series index were recorded in detail to facilitate subsequent vector construction. Based on the fermentation differential signal units, a differential representation vector set within the microbial tribe was constructed. The vector dimensions were defined as metabolic response delay time (in minutes), product conversion offset percentage (in%), and growth curve drift index. For each signal unit, the specific values ​​of these three indicators were extracted and normalized. The metabolic response delay time was mapped to the range 0 to 1 based on the maximum observed delay of 60 minutes, the product conversion offset was directly mapped to the range 0 to 1 based on the percentage, and the growth curve drift index was normalized to the observed maximum value of 0.5, forming a three-dimensional representation vector. This vector set was stored in matrix form, with the number of rows corresponding to the number of signal units and the number of columns fixed to 3. The matrix data was managed using the Pandas library in Python to ensure standardized data structure and scalability. The differential representation vector set was differentially labeled according to response intensity, starting offset point, and stable time period. The response intensity was defined as the weighted sum of the normalized values ​​of the three indicators, with weighting coefficients set to 0.4, 0.35, and 0, respectively.25. The starting offset point is determined by the actual time point corresponding to the metabolic response delay. The stable period is obtained by calculating the duration that the indicator in the time series remains above 80% of the peak value. The labeling operation uses a multidimensional thresholding method, dividing the response intensity into three levels: high, medium, and low. The starting offset point is divided into three time intervals: early, medium, and late. The stable period is divided into three segments: short, medium, and long. The label encoding is stored as a triplet, with a specific format such as (high, early, long). The differential labeling results are stored in JSON format to facilitate subsequent analysis by the automatic control system and the construction of microbial tribal fermentation differential data. The differential dataset contains label information and corresponding differential representation vectors.

[0094] Of particular importance is the construction of differential representation vector sets within microbial tribes based on fermentation differential signal units, including:

[0095] extracting differential signal feature data from fermentation differential signal units;

[0096] Normalizing the difference signal feature data to generate a normalized difference feature;

[0097] Perform cluster mapping on the normalized difference features to obtain signal unit clustering data;

[0098] Response trajectory coding was performed based on signal unit clustering data to generate within-tribe difference coding data;

[0099] The differentially encoded data within the tribe are vectorized and reorganized to obtain a differential representation vector set within the microbial tribe.

[0100] In this embodiment, based on the fermentation difference signal units obtained by cross-comparison screening of the previous time series, three core characteristic values ​​including metabolic response delay time, product conversion offset amplitude, and bacterial growth curve drift index are collected in detail for each unit. The metabolic response delay time refers to the time difference between the peak response time of the key metabolic indicator in the unit and the standard response time, and the unit is minutes. The product conversion offset amplitude is calculated by comparing the peak position and intensity difference of the metabolite concentration curve, and the unit is the percentage change of concentration. The bacterial growth curve drift index is quantified by fitting the exponential model parameter difference of its growth curve, specifically the exponential growth rate and the starting time offset. The Pandas data structure of Python is used to integrate the characteristic data in unit order to form a characteristic data table to ensure that the three-dimensional characteristic data of each signal unit are complete and corresponding. All characteristic values ​​are from the high-frequency time series data collected in the experiment, and the sampling interval is accurate to 5 minutes. Signal feature data were normalized using linear normalization, mapping each of the three features to the range 0 to 1. The normalization formula was (current value - minimum value) / (maximum value - minimum value). The metabolic response delay time had a maximum of 60 minutes and a minimum of 0 minutes, the product conversion offset amplitude had a maximum of 30% and a minimum of 0%, and the bacterial growth curve drift index had a maximum of 0.5 and a minimum of 0. Normalized values ​​were rounded to four decimal places to ensure accuracy. Batch normalization was performed on the data array using the Python Numpy library. All input data were floating-point. A normalized feature matrix was generated, with each row corresponding to a signal unit and each column corresponding to the normalized feature dimension. Normalization ensured uniform dimensionality across features. Cluster mapping was performed on the normalized differential features using the density-based spatial clustering of applications with noise (DBSCAN) algorithm, with the neighborhood radius parameter eps set to 0.1, the minimum number of samples min_samples is 5, the algorithm performs a spatial distribution scan on the normalized feature matrix, identifies dense signal units in the cluster, and ignores noise points. The clustering process is completed by calculating the Euclidean distance between each point in the feature space. The clustering result is output as the cluster label corresponding to each signal unit. The label value is a non-negative integer, -1 represents a noise point, and the clustering result is stored in a Python dictionary structure. The key is the signal unit number and the value is the cluster label. During the cluster mapping process, a three-dimensional scatter plot is drawn using the visualization tool Matplotlib, and the cluster labels are color-coded. The response trajectory encoding operation is performed based on the signal unit clustering data to encode the cluster labels. The encoding rule assigns binary codes in sequence according to the numerical order of the cluster labels. The encoding length is dynamically determined according to the number of clusters. The encoding generation follows the fixed prefix code principle to ensure that the encoding is unique and non-overlapping. The response trajectory encoding includes the label continuity information of continuous signal units in the time series, using A Python dictionary maps signal unit time series indices to their corresponding codes. The code generation algorithm is implemented based on a finite state machine model. By progressively scanning the time series unit indices, adjacent units with the same label are merged into continuous code segments. The resulting code data is stored as a string array, with string elements representing consecutive code segments and their start and end time points. The differentially coded data within a microbial community is vectorized and reorganized. The vector dimension is first defined as the number of code segments multiplied by the code length. The code segments are arranged in chronological order. The code strings are converted to a numerical matrix using ASCII code. After conversion, each row of the matrix corresponds to a code segment, and each column to a code bit. The matrix is ​​reorganized into a high-dimensional vector by row concatenation. The sparse matrix storage method of the Python Scipy library is used to reduce memory usage. The vectorized data is stored in HDF5 format to facilitate large-scale data access. The resulting differential representation vector set serves as a unified numerical representation within the microbial community.

[0101] Preferably, in step S5, performing fermentation difference compensation simulation on the simulated microbial environment field based on the fermentation difference data includes:

[0102] Mapping the fermentation difference data space to the simulated microbial environment field to obtain the microbial fermentation difference area;

[0103] Determine the compensatory intervention parameters required to compensate for the differences in microbial fermentation areas;

[0104] Based on the compensation intervention parameters, the virtual fermentation path of the simulated microbial environment field is deduced to obtain the simulation path data;

[0105] Correcting an error reverse drift in the simulation path data according to the fermentation difference data to generate error drift correction data;

[0106] A compensation simulation is performed based on the error drift correction data to obtain simulated compensation data.

[0107] In this embodiment, the fermentation difference data space is mapped to the operation of simulating the microbial environment field. First, a three-dimensional microbial environment field model is established. The model size is determined according to the internal structure of the actual fermentation tank. The length, width and height are set to 2 meters, 1 meter and 1 meter respectively. It is divided into grid units of equal volume, and the unit side length is set to 5 centimeters. The spatial coordinates of each microbial tribe in the previously obtained fermentation difference data and its difference characterization vector are used to map them to the corresponding grid units in the environmental field. The mapping adopts a weighted distribution algorithm. The weight is determined according to the response intensity of the difference characterization vector. The mapping result forms a difference density distribution matrix. Each unit in the matrix contains a difference intensity value. NumPy in Python is used for matrix operations to realize data from difference characterization. The conversion of vectors to environmental field space grids ensures that the accuracy of spatial mapping reaches a spatial resolution of 5 cm. The mapping data provides the basis for subsequent regional division and compensation parameter calculation. The compensation intervention parameters required for difference compensation are determined by the microbial fermentation difference area. The threshold segmentation method is used to partition the difference density distribution matrix. The grid cells with difference intensity values ​​greater than 0.7 are divided into high difference areas, those between 0.3 and 0.7 are divided into medium difference areas, and those below 0.3 are divided into low difference areas. According to the regional division, combined with the microbial community structure characteristics and fermentation process parameters, the corresponding compensation intervention parameters are set, including the dissolved oxygen concentration adjustment range (unit: mg / L), which is controlled between 5 and 10 mg / L, and temperature compensation. The temperature range (unit: degrees Celsius) is controlled within the range of 30 to 37 degrees Celsius, and the stirring speed adjustment range (unit: rpm) is controlled between 150 and 250 rpm. The statistical analysis software R is used for parameter aggregation and optimization. Combined with the microbial community activity data in the region, a difference compensation parameter database is generated to support virtual fermentation path deduction. Based on the compensation intervention parameters, the virtual fermentation path deduction of the simulated microbial environment field is carried out. First, the compensation parameters are input into the multi-physics simulation software COMSOL Multiphysics to construct a coupling model including fluid dynamics, heat conduction and microbial metabolism. The boundary conditions are set to the actual operating parameters of the fermentation tank, the time step is set to 1 minute, and the total simulation time is 72 hours. During the simulation, the dissolved oxygen, temperature and stirring speed in the grid were dynamically adjusted. The simulation path data included changes in microbial community density, metabolite concentration and pH value. The simulation data was output in time series data format. The file type was CSV, which contained time series response data of each grid unit. The simulation data was post-processed using Python scripts to extract key path trajectories for subsequent error correction. The error reverse drift in the simulation path data was corrected according to the fermentation difference data. The error reverse drift was identified by comparing the time series of key parameters in the simulation path with the corresponding time point parameters in the fermentation difference data, and calculating the error deviation. The error was defined as the simulation value minus the actual fermentation measurement value. Dynamic Time Warping (DTW) was used.The DTW) algorithm is used to align the error time series, identify the drift trend of the error over time, set the drift threshold to the time period when the absolute value of the simulation error is greater than 5%, and perform reverse correction on the drift interval. The correction process constructs an error compensation function and fits the error drift with time as the independent variable. The fitting function type adopts a polynomial regression model with a third-order order. After the fitting function is generated, its value is reversely applied to the simulation path data to realize dynamic error correction. The corrected data is stored in matrix form. Compensation simulation is performed based on the error drift correction data. The compensation simulation process adopts a numerical iteration method. The corrected simulation path data and compensation intervention parameters are input and the finite difference is used. The method iteratively updates the state variables of each grid cell in the environmental field. The state variables include microbial concentration, metabolite concentration, and environmental factors. During the iterative process, each update is based on the weighted average of the correction factor of the calibration data and the compensation parameter. The weight distribution is optimized using a genetic algorithm. The optimization objective function is to minimize the error between the simulation data and the experimental observation data. The number of iterations is set to 1000. The simulation results generate a multidimensional array with the array dimension being the number of spatial grids multiplied by the number of time steps multiplied by the number of variables. The simulation compensation data is stored in HDF5 format to achieve efficient access to large-scale data. The compensation simulation results serve as input parameters for the subsequent automatic control of the fermentation process, enabling refined control and adjustment.

[0108] Preferably, performing automatic fermentation control according to the simulated compensation data in step S5 includes:

[0109] Performing control variable mapping according to the simulation compensation data to obtain control variable data;

[0110] Perform synchronization and coordination of microbial environment based on control variable data to generate synchronization and coordination data for fermentation control;

[0111] The automatic control framework of microbial reaction fermentation is constructed by synchronously coordinating data of fermentation control to carry out automatic control of microbial reaction fermentation.

[0112] In this embodiment, the control variable mapping operation is performed according to the simulation compensation data. First, the simulation compensation data is read. The data is stored in the form of a multidimensional matrix and includes key variables such as time series temperature, dissolved oxygen concentration, pH value and microbial concentration. The fermentation environment variables in the simulation compensation data are converted into control variable values ​​using a data mapping algorithm. The control variables include heating power (unit: watt), aeration flow (unit: liters per minute), and stirring speed (unit: revolutions per minute). The mapping rules are constructed based on a multivariate linear regression model. The model input is the simulation compensation temperature, oxygen concentration and pH data, and the output is the corresponding control variable value. The regression model parameters are obtained through historical fermentation process data. According to the training, the input data is normalized. The normalization method uses minimum-maximum scaling to scale all input variables to the range of 0 to 1. The weighted adjustment mechanism is applied in the mapping process. The weights are set according to the sensitivity of different control variables to the fermentation performance. The mapping calculation is implemented using the scikit-learn library in Python. After the mapping is completed, a time series control variable data set is generated to provide basic data for subsequent microbial environment synchronization and coordination. The implementation operation of microbial environment synchronization and coordination is carried out based on the control variable data. First, the control variable data is divided into control action sequences of different time periods. The time period length is set to 10 minutes. The sequence includes heating power, exposure, and so on. The three parameters of air flow and stirring speed are the air flow rate and stirring speed. A real-time synchronization coordination module is set for different parameters. A control strategy based on multi-agent collaboration is adopted. Each control variable is regarded as an agent. The agents coordinate by sharing state information. The state information includes the current fermentation tank temperature, dissolved oxygen concentration and microbial community activity index. The Q-learning method in the reinforcement learning algorithm is used to train the collaborative model. The state space is defined as the interval after the above environmental parameters are discretized, and the action space is defined as the discrete step size of the control variable adjustment. The reward function is designed to minimize the difference between the environmental parameter deviation and the set target. The simulation data enhancement strategy is adopted in the training process. The previously generated simulation compensation data is used as the training environment input. After the training is completed, the synchronization coordination module continuously adjusts the control variables through state monitoring in real time. The output is fermentation control synchronization coordination data. The data structure is a multidimensional time series matrix. The matrix contains the real-time adjustment instructions and corresponding timestamps of each control variable, realizing dynamic synchronization coordination of the microbial environment. The fermentation reaction fermentation is automatically controlled by the fermentation control synchronization coordination data. First, the fermentation control synchronization coordination data is input into the fermentation control execution system, which includes a heater control unit, an aeration system adjustment unit and a stirring motor drive unit. The control system adopts an industrial field bus (Industrial Field Bus) based control system. Fieldbus) real-time communication protocol to ensure that control instructions are delivered to each execution device within milliseconds, and the control signal format complies with the Controller Area Network (Controller Area Network,The control execution unit continuously monitors environmental parameter data collected by temperature, dissolved oxygen, and pH sensors through a real-time feedback acquisition module. The feedback data is compared with the synchronized coordination data to form a closed-loop control mechanism. The control system sets threshold errors of ±0.1 degrees Celsius for temperature, ±0.2 mg / L for dissolved oxygen, and ±0.05 for pH. When the thresholds are exceeded, the control variable mapping and environmental synchronization are automatically recalculated, and the control parameters are automatically adjusted. The control framework integrates a software architecture based on a distributed control system (DCS), running multiple parallel control threads to achieve multi-dimensional microbial reaction and fermentation control. The state monitoring module enables real-time dynamic adjustment of the fermentation process, ensuring that fermentation process parameters are always maintained within the preset range, completing the construction of the microbial reaction and fermentation automatic control framework.

[0113] The present invention also provides a microbial reaction and fermentation automatic control system for executing the microbial reaction and fermentation automatic control method described above. The microbial reaction and fermentation automatic control system comprises:

[0114] The parameter extraction module is used to collect real-time microbial environmental parameters; identify the multi-axis structure of the real-time microbial environmental parameters and extract the main control environment interval data based on the multi-axis structure; eliminate redundant signals from the main control environment interval data to generate standard microbial environmental data;

[0115] A fermentation mapping module is used to define a fermentation degree control template based on fermentation history record data; construct a microbial fermentation timeline framework based on the fermentation degree control template; and identify the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework;

[0116] The simulation field building module is used to build a simulated microbial environment field using standard microbial environment data; based on the microbial fermentation state and the simulated microbial environment field, the global microbial reaction simulation is performed on the standard microbial environment data to generate global simulated microbial reaction data;

[0117] The difference recognition module is used to identify the microbial community structure in the simulated microbial environment field and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data. The global simulated microbial reaction data is compared with each partition according to the fermentation reaction structure partition to generate fermentation difference data within the microbial tribe.

[0118] The intelligent compensation module is used to simulate the fermentation difference compensation of the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; and automatically control the fermentation according to the simulated compensation data, thereby realizing automatic fermentation compensation of microorganisms.

[0119] The present invention realizes the efficient collection and analysis of real-time microbial environmental parameters through the implementation of the parameter extraction module, can accurately identify the multi-axis structure of data and extract the main control environment interval data, and the generated standard microbial environmental data ensures the reliability and effectiveness of subsequent processing. The introduction of the fermentation mapping module enables the effective use of fermentation history record data. The defined fermentation degree control template provides a scientific basis for the judgment of the microbial state. The constructed microbial fermentation timeline framework enhances the understanding and grasp of the dynamic changes of the fermentation process. The simulation field building module builds a simulated microbial environment field and performs global microbial reaction simulation based on standard microbial environment data, realizing a more comprehensive reaction characteristic analysis. The generated global simulated microbial reaction data provides a solid foundation for subsequent difference identification. The difference identification module The structural partitioning of local simulated microbial reaction data ensures the accurate understanding and control of each reaction area. The generated fermentation difference data within the microbial tribe reveals the key differences in the fermentation process. Based on these difference data, the intelligent compensation module can timely perform compensation simulation on the simulated microbial environment field. The generated simulated compensation data provides a scientific basis for the automatic control of the fermentation process, ensuring that each fermentation area is carried out under optimal conditions, and ultimately realizes the automatic compensation and optimization of the microbial fermentation process, improves fermentation efficiency, and reduces the need for manual intervention. The realization of the overall system not only improves the stability and consistency of the microbial fermentation process, reduces production costs, and improves resource utilization, but also promotes the intelligent and automated process of microbial fermentation technology, and meets the modern industry's demand for efficient and precise production.

[0120] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0121] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A microbial reaction fermentation automatic control method, characterized in that: The following steps are involved: Step S1: collecting real-time microbial environmental parameters; identifying the multi-axis structure of the data of the real-time microbial environmental parameters, and extracting the main control environment interval data based on the multi-axis structure of the data; eliminating redundant signals through the main control environment interval data, thereby generating standard microbial environmental data; Step S2: defining a fermentation degree control template based on the fermentation history record data; constructing a microbial fermentation timeline framework based on the fermentation degree control template; and identifying the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework; Step S3: constructing a simulated microbial environment field using standard microbial environment data; performing a global microbial reaction simulation on the standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field, thereby generating global simulated microbial reaction data; Step S4: Identify the microbial community structure in the simulated microbial environment field, and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data; compare the global simulated microbial reaction data for each partition based on the fermentation reaction structure partition, thereby generating fermentation difference data within the microbial tribe; Step S5: performing fermentation difference compensation simulation on the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; performing automatic fermentation control according to the simulated compensation data, thereby realizing automatic microbial fermentation compensation.

2. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: integrating the original data of temperature, pH value, dissolved oxygen, and pressure collected by the sensor into real-time microbial environmental parameters; Step S12: performing signal mean level conversion on the real-time microbial environmental parameters to obtain a multi-axis conversion matrix, wherein the conversion step range is 0.2-0.5 units; Step S13: performing difference convergence filtering on the multi-axis conversion matrix to generate the main control section stability parameter, wherein the filter window width range is 5-15 groups of data, and the filter threshold is that the standard deviation of each axis does not exceed 0.35 units; Step S14: extracting the main control environment interval data from the main control section stability parameters, wherein the time span of the main control section is not less than 20 minutes; Step S15: determining the mutation environment parameters in the real-time microbial environment parameters based on the main control environment interval data; Step S16: Eliminate redundant signals in the real-time microbial environmental parameters based on the mutant environmental parameters, thereby generating standard microbial environmental data.

3. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: querying the fermentation history data of the microorganism through the log; mapping the fermentation process characteristic intervals according to the fermentation history data to define a fermentation degree control template, wherein the fermentation history data has a temperature range of 28-38°C, a pH range of 5.5-7.2, a dissolved oxygen range of 15-85%, and a pressure range of 0.8-1.2 bar; Step S22: Synchronously pairing the fermentation degree control template with the stage anchor points to obtain timeline pairing anchor points; constructing a microbial fermentation timeline framework through the timeline pairing anchor points, wherein the time interval corresponding to each stage in the microbial fermentation timeline framework ranges from 8 to 30 hours, and the state anchor point matching deviation is controlled within a range of 2 hours; Step S23: projecting the fermentation characteristic data in the standard microbial environment data into the microbial fermentation time frame, and performing fermentation characteristic pairing to obtain paired fermentation characteristics; Step S24: determining the corresponding microbial fermentation state in the microbial fermentation timeline frame based on the paired fermentation characteristics.

4. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: In step S3, the construction of a simulated microbial environment field using standard microbial environment data includes: Extract various environmental characteristic parameters from standard microbial environmental data; Reconstruct the spatial grid environment from standard microbial environmental data according to environmental characteristic parameters; Identify local heterogeneous features in a spatial grid environment; Correcting local heterogeneous features based on environmental feature parameters to generate corrected environmental features; Construct a simulated microbial environment field by correcting environmental characteristics.

5. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: In step S3, the global microbial reaction simulation of the standard microbial environment data based on the microbial fermentation state and the simulated microbial environment field includes: Projecting the microbial fermentation state onto the simulated microbial environment field to obtain the fusion fermentation environment state; Mapping the fusion fermentation environment state to the reaction tendency field to obtain reaction tendency field data; Retrieve local microbial fermentation diffusion paths by comparing reaction tendency field data with standard microbial environment data; Build collaborative feedback links based on local microbial fermentation and diffusion pathways; The spatiotemporal evolution driving is performed according to the collaborative feedback link, thereby obtaining the spatiotemporal evolution driving data; The local microbial responses are simulated by driving data based on spatiotemporal evolution, and then superimposed and integrated into global simulated microbial response data.

6. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: In step S4, identifying the microbial community structure in the simulated microbial environment field and determining the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data includes: Hierarchical identification of microbial population activity trajectories in simulated microbial environment fields; Perform temporal clustering of microbial population activity trajectories to generate multiple microbial subpopulation clustering areas; The community structure was analyzed by microbial subpopulation aggregation areas to determine the microbial community structure; The global simulated microbial reaction data is divided into multi-dimensional grids to obtain the microecological activity heat, metabolite accumulation rate and growth curve transfer rate; A three-axis indicator mapping matrix was constructed based on the microecological activity heat, metabolite accumulation rate, and growth curve transfer rate; The sensitivity of different microbial subpopulation clusters was quantified based on the three-axis index mapping matrix, and the dynamic fermentation response diagram of each local microbial unit was drawn; The microbial community structure was spatially overlaid with the dynamic fermentation response map, and high response trajectory areas were identified; The standard microbial environment data are divided into multiple fermentation reaction structure partitions based on the high response trajectory areas.

7. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: In step S4, the global simulated microbial reaction data is partitioned according to the fermentation reaction structure, and the partition differences are compared, including: Perform sequence cross-comparison on the microbial response time series within the fermentation reaction structure partition to obtain partition response cross-data; Extracting fermentation difference signal units that exhibit any of the following phenomena: metabolic response delay, product conversion deviation, and growth curve drift from the partition response cross-data; Based on the fermentation difference signal unit, a differential representation vector set within the microbial tribe was constructed; The differential characterization vector set is differentially labeled according to the response intensity, starting offset point and stable time period, thereby generating fermentation difference data within the microbial tribe.

8. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: The fermentation difference compensation simulation of the simulated microbial environment field based on the fermentation difference data in step S5 includes: Mapping the fermentation difference data space to the simulated microbial environment field to obtain the microbial fermentation difference area; Determine the compensatory intervention parameters required to compensate for the differences in microbial fermentation areas; Based on the compensation intervention parameters, the virtual fermentation path of the simulated microbial environment field is deduced to obtain the simulation path data; Correcting an error reverse drift in the simulation path data according to the fermentation difference data to generate error drift correction data; A compensation simulation is performed based on the error drift correction data to obtain simulated compensation data.

9. The microbial reaction fermentation automatic control method according to claim 1, characterized in that: The automatic fermentation control according to the simulated compensation data in step S5 includes: Performing control variable mapping according to the simulation compensation data to obtain control variable data; Perform synchronization and coordination of microbial environment based on control variable data to generate synchronization and coordination data for fermentation control; The automatic control framework of microbial reaction fermentation is constructed by synchronously coordinating data of fermentation control to carry out automatic control of microbial reaction fermentation.

10. A microbial reaction and fermentation automatic control system, characterized in that: For executing the microbial reaction and fermentation automatic control method according to claim 1, the microbial reaction and fermentation automatic control system comprises: The parameter extraction module is used to collect real-time microbial environmental parameters; identify the multi-axis structure of the real-time microbial environmental parameters and extract the main control environment interval data based on the multi-axis structure; eliminate redundant signals from the main control environment interval data to generate standard microbial environmental data; A fermentation mapping module is used to define a fermentation degree control template based on fermentation history record data; construct a microbial fermentation timeline framework based on the fermentation degree control template; and identify the microbial fermentation state to which the standard microbial environment data belongs based on the microbial fermentation timeline framework; The simulation field building module is used to build a simulated microbial environment field using standard microbial environment data; based on the microbial fermentation state and the simulated microbial environment field, the global microbial reaction simulation is performed on the standard microbial environment data to generate global simulated microbial reaction data; The difference recognition module is used to identify the microbial community structure in the simulated microbial environment field and determine the fermentation reaction structure partition based on the microbial community structure and the global simulated microbial reaction data. The global simulated microbial reaction data is compared with each partition according to the fermentation reaction structure partition to generate fermentation difference data within the microbial tribe. The intelligent compensation module is used to simulate the fermentation difference compensation of the simulated microbial environment field based on the fermentation difference data, thereby obtaining simulated compensation data; and automatically control the fermentation according to the simulated compensation data, thereby realizing automatic fermentation compensation of microorganisms.

Images