Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on online Raman spectrum

Abstract

The invention provides a Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on an online Raman spectrum, belongs to the technical field of biological fermentation process control, and aims to solve the problems of unstable process and low efficiency caused by fermentation control lag and incapability of sensing the real metabolic state of cells in the prior art. The method comprises the following steps: acquiring the concentrations of a target product BDO and key byproducts such as acetic acid and ethanol in the fermentation liquor in real time through an online Raman spectrum; according to the method, a metabolic stress index is originally proposed and constructed, the index is obtained by performing weighted operation on the instantaneous generation rate of the by-product and the target product, and the index is used for quantitatively characterizing the intrinsic metabolic stress level of the cells in real time. The control strategy of maintaining the metabolic stress index in the preset optimal stable interval is taken as a core control strategy, the conversion from passive response to active prediction in the fermentation process is realized, and the yield, the stability and the batch repeatability of Bi-BDO production are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biological chemical fermentation process control, in particular to a Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on online Raman spectrum. BACKGROUND

[0002] BDO is an important platform chemical, which is widely used in the production of engineering plastics, polyurethane, solvents and drugs. The traditional production of BDO highly depends on petrochemical route, which faces the problems of resource consumption and environmental pollution. With the development of synthetic biology and metabolic engineering technology, the biological route of producing BDO through microbial fermentation has become a research hotspot and an important direction of future industrialization in this field, because it has the advantages of renewable raw materials and green and environmentally friendly process.

[0003] The core of producing BDO by biological method is the microbial fermentation process. The efficiency of the process, including product yield, production intensity and final concentration, directly determines its economic feasibility. In the fermentation process, pH and dissolved oxygen (DO) are the two most critical environmental parameters that affect microbial growth, substrate consumption and product synthesis metabolic pathways. Therefore, accurate and effective control of pH and dissolved oxygen is the prerequisite for efficient and stable production of BDO.

[0004] Currently, the control of BDO fermentation process mainly has the following technical problems: 1. The control method based on offline sampling analysis has serious hysteresis and data sparsity problems: Traditional fermentation process monitoring relies on aseptic sampling from the fermentation tank at regular intervals, and then using high-performance liquid chromatography and other analytical instruments to detect the concentrations of substrates, products and byproducts in the sample. This method has fundamental defects: first, the entire analysis process takes several hours, and by the time the detection results show that the process is abnormal, the metabolic state of the fermentation system has already deviated irreversibly, and the intervention of control measures is too late to recover the loss. Secondly, to reduce the risk of pollution, the sampling frequency is low, resulting in sparse and discontinuous process data, and the control personnel cannot grasp the dynamic changes of the fermentation process during the sampling interval.

[0005] 2. The constant value control strategy based on online probe is out of touch with the dynamic physiological needs of cells: To overcome the hysteresis of offline analysis, online pH electrode and dissolved oxygen electrode are generally used in industrial fermentation to maintain these parameters at a constant value set in advance. However, the fundamental problem of this control strategy is that the control target is the apparent physical and chemical parameters, not the real internal physiological state of the cells. Microorganisms have different optimal requirements for the environment in different stages of fermentation. In the rapid growth period, cells need higher dissolved oxygen to support rapid proliferation; and in the product synthesis period, micro-aerobic or anaerobic conditions are needed to optimize the metabolic flow to the target product. The rigid maintenance of pH and dissolved oxygen at a fixed value cannot adapt to the dynamic metabolic demand of cells in the entire fermentation cycle, thereby limiting the maximum production potential of the strain in most of the time.

[0006] 3. The existing online monitoring technology fails to solve the problem of the superficiality of the control logic: In recent years, process analysis technologies represented by online Raman spectroscopy have been applied to fermentation processes, which can obtain the concentration information of multiple components in the fermentation broth in real time, which to some extent solves the problem of data hysteresis. However, the current control logic based on these online data still stays at a relatively superficial level, which adjusts the stirring speed or feed rate by monitoring the real-time generation rate of BDO. This control method is essentially a post-response, which adjusts after observing the result of the decrease in production rate, and does not touch the root cause of the rate decrease. This method cannot understand the metabolic network state inside the cells, especially cannot perceive the internal driving force that causes the metabolic flow to shift from the BDO synthesis path to the byproduct path, and the metabolic stress state such as redox imbalance inside the cells. Therefore, its control effect is still indirect and hysteresis, and cannot realize the prospective guidance and active maintenance of the metabolic path.

[0007] In summary, the existing BDO fermentation control technology generally lacks the perception of the internal physiological state of the cells, which leads to the limitation that the control strategy based on offline data or online data is difficult to break away from passive response and blind maintenance, and cannot fundamentally guarantee the continuous and efficient guidance of metabolic flow to the target product. This has become the core technical bottleneck restricting the industrialization efficiency and stability of Bi-BDO technology. SUMMARY

[0008] Technical problems solved In view of the shortcomings of the prior art, the present application provides a Bi-BDO fermentation pH and dissolved oxygen dynamic optimization method based on online Raman spectroscopy, which solves the following problems: 1. The problem of serious lag and data blind spot in traditional fermentation control is solved: traditional control relies on offline sampling and HPLC analysis, and there is a time lag of several hours. When abnormal accumulation of by-products is detected, the cell metabolism has already deviated irreversibly, and the control measures can only be passive remediation, missing the best intervention window. At the same time, low-frequency sampling results in sparse process data, forming a large number of monitoring blind spots, which cannot capture key dynamic changes, leading to a high risk of process out-of-control.

[0009] 2. The problem of serious disconnection between control target and real metabolic state of cells is solved: existing control strategies generally maintain pH, dissolved oxygen and other parameters at constant values. However, the physiological needs of microorganisms are dynamically changing at different growth and production stages. This method of using apparent physical and chemical parameters as control targets is completely disconnected from the intrinsic and dynamic metabolic network state, and cannot always provide the optimal growth and production environment, thereby severely limiting the full play of the production potential of the strain.

[0010] 3. The problem of shallow control logic and lack of forward-looking predictive ability is solved: even with online concentration monitoring, existing control logic still stays at the passive response to the result indicators such as product generation rate. It cannot understand the root cause of the decline in production efficiency, i.e. the metabolic stress state inside the cell. Due to the lack of sensing ability for the precursors of metabolic imbalance, the control system cannot make real forward-looking intervention, but can only wait for the problem to occur and then make corrections. The control level is shallow and cannot guarantee the stability of the process from the source.

[0011] Technical scheme To achieve the above object, the present application is implemented by the following technical scheme: a Bi-BDO fermentation pH and dissolved oxygen dynamic optimization method based on online Raman spectrum, comprising the following steps: Sp1. Online spectrum and concentration data acquisition: during the BDO fermentation process, online Raman spectrum data of the fermentation broth are continuously collected, and concentration time series data of the target product BDO and by-products related to cell metabolic stress are obtained in real time by chemometrics model analysis; Sp2. Metabolic stress state quantification: based on the concentration time series data obtained in Sp1, the first derivative is taken to obtain the instantaneous generation rate of the target product and the by-products, and a comprehensive index for real-time quantification of the overall metabolic stress level of the fermentation system is calculated according to a predetermined function relationship, defined as metabolic stress index; Sp3. Metabolic imbalance trend prediction: analyze the time series of metabolic stress index calculated in Sp2 to obtain its current value and time change gradient. When the value of the metabolic stress index exceeds the preset baseline threshold, or its time change gradient shows a sustained positive growth trend, it is determined that there is a precursor of metabolic imbalance in the fermentation process; Sp4. Adjusting decision and execution: when Sp3 determines the existence of metabolic imbalance precursors, the preset adjustment logic is started to apply a small amplitude adjustment to the pH and dissolved oxygen control variables, and returns to Sp1 to continue monitoring the inhibitory effect of the adjustment on the metabolic stress index; Sp5. Closed-loop optimization: according to the actual inhibitory effect of the adjustment on the metabolic stress index in Sp4, the direction and amplitude of subsequent adjustment are adaptively adjusted to optimize and maintain the metabolic stress index in the preset optimal interval through dynamic adjustment of pH and dissolved oxygen, so as to realize dynamic optimization of the fermentation process.

[0012] Preferably, the by-product related to cell metabolic stress in Sp1 is one or a combination of acetic acid, ethanol, lactic acid or pyruvic acid.

[0013] Preferably, the weighting coefficient for weighting the instantaneous generation rates of by-products and target product BDO to calculate the metabolic stress index in Sp2 is obtained by statistical analysis and optimization calculation of historical fermentation batch data under different pH and dissolved oxygen conditions, and the optimization target is to maximize the negative correlation between the metabolic stress index and the final BDO yield.

[0014] Preferably, the baseline threshold and the optimal interval of the metabolic stress index in Sp3 are determined by statistical analysis of the metabolic stress index data of high-yield stable fermentation batches.

[0015] Preferably, the adjustment logic in Sp4 is a hierarchical response logic: when it is determined that there is a slight imbalance precursor, only the dissolved oxygen control variable is fine-tuned; when it is determined that there is a significant imbalance precursor, the dissolved oxygen and pH control variables are adjusted cooperatively.

[0016] Preferably, the adjustment amount in Sp4 is a small amplitude adjustment of 0.1% to 5% of the current set value.

[0017] Preferably, the closed-loop optimization in Sp5 is an adaptive feedback optimization that uses a hill climbing algorithm or a particle swarm optimization algorithm to minimize the metabolic stress index and search for the optimal pH and dissolved oxygen adjustment amount online.

[0018] Preferably, a system for Bi-BDO fermentation pH and dissolved oxygen dynamic optimization based on online Raman spectroscopy includes: Bi-BDO bioreactor equipped with pH and dissolved oxygen sensing and execution mechanism; Online Raman spectroscopy analyzer; Central processing unit configured to: Receive and analyze Raman spectroscopy data to obtain concentration time series; According to Sp2, the metabolic stress index is calculated in real time; According to Sp3, the metabolic imbalance trend is predicted in real time; According to Sp4 and Sp5, adjustment instructions are generated and sent to the actuators of the fermenter to form a closed-loop control.

[0019] Preferably, the central processor further comprises a historical database and a model optimization module for storing historical fermentation data and for offline optimization and updating of the weighting coefficients used in calculating the metabolic stress index.

[0020] Advantages

[0021] The present application provides a Bi-BDO fermentation pH and dissolved oxygen dynamic optimization method based on online Raman spectrum. It has the following advantages: 1. The present application uses online Raman spectrum to continuously acquire data and calculates the metabolic stress index that can represent the internal state of cells in real time, providing a continuous and high-resolution process dashboard for the control system. This fundamentally solves the serious lag problem of offline sampling analysis lasting for several hours, as well as the data sparsity and process blind spot problem caused by low-frequency sampling. The control system can instantaneously perceive the stress state fluctuations caused by subtle changes in metabolic flow, so that the basis for decision-making and operation is no longer outdated information from several hours ago, but the real-time and dynamic process state, realizing the leap from lag response to real-time insight and completely overcoming the inertia of traditional control methods.

[0022] 2. The existing technology either maintains constant pH / DO values or tracks changes in product concentration, with the control target being an external and indirect apparent parameter. The present application takes a unique approach by directly setting the control target to maintain the metabolic stress index in the optimal stable interval. This index is a direct quantitative representation of the redox balance and energy state of cells and is the central determinant of metabolic flow direction. By directly controlling this core physiological indicator, the present application ensures that every adjustment of pH and dissolved oxygen precisely serves the fundamental purpose of stabilizing the internal metabolic network of cells, making environmental control and dynamic physiological needs of cells always synchronized and optimally matched, thereby maximizing the production potential of strains throughout the fermentation cycle and realizing a profound change in control targets from apparent parameters to core physiological states, ensuring the precision and efficiency of control.

[0023] 3、The most core advantage of the present application lies in its foresight. By monitoring the metabolic stress index and its change gradient, the method can sensitively capture the precursor of imbalance before the large-scale, irreversible deviation of the cell metabolic network occurs. It occupies the valuable window period between the deviation of metabolic trend and the deterioration of production performance, and actively pulls the cell back to the healthy state of high-efficiency BDO production through the implementation of minimal and timely perturbation adjustment. This completely changes the situation of passive remediation after detecting the malignant results such as yield reduction or by-product accumulation in the prior art, realizes the prevention of the fermentation process, avoids the toxic inhibition caused by the invalid consumption of carbon source and by-product accumulation from the source, and realizes the fundamental change of the control logic from passive remediation to active foresight and prospective stability.

[0024] 4、Since the present application can continuously maintain the cell in the optimal metabolic interval of low stress and high activity, it effectively inhibits various metabolic unexpected turns that lead to process fluctuation and efficiency reduction. Therefore, the application of the present application can significantly improve the final yield and production intensity of BDO, and greatly reduce the generation of impurities such as acetic acid and ethanol. More importantly, by converting the uncertain and easily drifting cell physiological state into a quantifiable and controllable index, the present application greatly enhances the stability and predictability of the fermentation process, thereby significantly improving the batch-to-batch repeatability of production, providing a key technical support for the large-scale and stable industrial production of Bi-BDO technology, and significantly improving the stability, yield and batch-to-batch repeatability of the BDO fermentation process. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The system composition cloud chart of the present application; Figure 2 The system architecture diagram of the present application; Figure 3 The work flow chart of the present application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application. Specific embodiment one:

[0028] As shown in Figures 1-2 A Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on online Raman spectrum, comprising the following steps: Sp1. Online spectrum and concentration data acquisition: During the BDO fermentation process, the online Raman spectrum data of the fermentation broth is continuously collected, and the concentration time series data of the target product BDO and the by-products related to cell metabolic stress are obtained in real time by chemometrics model analysis; Sp2. Metabolic stress state quantification: Based on the concentration time series data obtained in Sp1, the first derivative is obtained to obtain the instantaneous generation rate of the target product and the by-products, and a comprehensive index for real-time quantification of the overall metabolic stress level of the fermentation system is calculated according to the preset function relationship, which is defined as the metabolic stress index; Sp3. Metabolic imbalance trend prediction: The time series of the metabolic stress index calculated in Sp2 is analyzed to obtain its current value and its time change gradient. When the value of the metabolic stress index exceeds the preset baseline threshold, or its time change gradient shows a sustained positive growth trend, it is determined that there is a precursor of metabolic imbalance in the fermentation process; Sp4. Adjustment decision and execution: When Sp3 determines that there is a precursor of metabolic imbalance, the preset adjustment logic is started to apply a small amount of adjustment to the pH and dissolved oxygen control variables, and returns to Sp1 to continuously monitor the inhibitory effect of the adjustment on the metabolic stress index; Sp5. Closed-loop optimization: According to the actual inhibitory effect of the adjustment amount in Sp4 on the metabolic stress index, the direction and amplitude of subsequent adjustment are adaptively adjusted, and the metabolic stress index is optimized and maintained in the preset optimal interval by dynamically adjusting pH and dissolved oxygen, realizing dynamic optimization of the fermentation process.

[0029] A Bi-BDO fermentation pH and dissolved oxygen dynamic optimization method based on online Raman spectrum: The core idea of this method is to construct a soft measurement index that can reflect the internal metabolic health status of microorganisms in real time and quantitatively through online Raman spectrum technology. This index is named metabolic stress index. Then, taking maintaining the index in the optimal stable interval as the control target, pH and dissolved oxygen are dynamically optimized and adjusted in advance, which realizes the fundamental guidance and stability of cell metabolic flow, and finally achieves efficient and stable Bi-BDO production process.

[0030] 1. Construction and connotation of metabolic stress index: The design principle of metabolic stress index is based on a deep understanding of the imbalance of microbial metabolic network. When the microbial cell is under environmental stress, the lack of dissolved oxygen in the fermentation broth leads to redox imbalance, or the glycolysis rate exceeds the processing capacity of the tricarboxylic acid cycle under high substrate concentration, leading to energy metabolism disorder. The cell will start the emergency metabolic pathway, and the generation of acetic acid is a typical marker of energy metabolism overflow, which is related to the ATP generation demand under high glycolysis flux. The generation of ethanol is a classic response to excessive reduction pressure, which is a temporary electron acceptor to regenerate oxidized coenzyme I, i.e. NAD+.

[0031] Therefore, the calculation of metabolic stress index is the difference between the total generation rate of stress byproducts after weighted processing and the generation rate of target products after weighted processing.

[0032] The stress byproducts here are specifically designated as acetic acid and ethanol, which have significant indicative significance for the metabolic network of Bi-BDO producing strain. The instantaneous generation rate of each stress byproduct is obtained by first-order derivative operation on the concentration data obtained by real-time analysis of Raman spectrum in time dimension. This rate value directly reflects the intensity of metabolic flux at this moment.

[0033] The weighting coefficients used for calculation are not empirically set, but have a solid data basis. They are obtained by strict multivariate nonlinear regression analysis or machine learning algorithms such as gradient boosting machine training on multiple batches of historical fermentation data. The data set analyzed includes process parameters, continuous concentration curves of each component, and final BDO yield and recovery rate of each batch. The goal of the optimization algorithm is to find a combination of weighting coefficients that can make the calculated metabolic stress index and the final BDO production performance indicators present the strongest negative correlation, so as to ensure the accuracy and reliability of the index as a cell metabolic health status indicator.

[0034] 2. Prospective dynamic optimization control: The control target of this method is not to track a certain fixed pH or dissolved oxygen set value, but to actively maintain the metabolic stress index in a pre-marked low stress optimal interval close to zero.

[0035] The core driving force of the control logic comes from the comprehensive analysis of the current value of the metabolic stress index and its time variation trend. The variation trend is quantified by calculating the second derivative of the time series of the index or the linear regression slope in a short time window. This mechanism enables the control system to predict the risk of metabolic imbalance through the slight positive drift trend of the metabolic stress index before the actual yield of BDO is substantially affected, and immediately start the adjustment program to achieve truly prospective control. Specific embodiment two:

[0037] AsFigures 1-2 As shown, a Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on online Raman spectrum, the running operation process: The complete operation process of the method is divided into two closely linked stages of offline modeling and system initialization and online dynamic optimization.

[0038] Stage one: offline modeling and system initialization; This stage is the basis for the successful application of the entire method, and is executed once before formal production.

[0039] Diversified fermentation data acquisition and accurate calibration: In order to ensure the robustness and wide applicability of the model, a series of BDO fermentation batches with diversity need to be designed and run, which should include high-yield stable-yield "golden batches", "abnormal batches" with large amounts of by-products due to reasons, and "exploratory batches" with deliberately changed pH and dissolved oxygen set points using experimental design method response surface method. During the running of all batches, two key operations are carried out simultaneously: first, the online Raman spectrum probe continuously collects spectral data at a high frequency; second, sterile sampling is strictly carried out according to the preset time point, and the concentrations of BDO, glucose, acetic acid, ethanol and other key components are immediately measured accurately using the calibrated high-performance liquid chromatograph. These offline measurement values will be used as the true value or label data for subsequent model training.

[0040] Construction and verification of chemometrics model: Using the collected spectral data and corresponding offline concentration true values, a mathematical model that can accurately predict the concentrations of multiple components from complex spectral signals is constructed using the partial least squares regression algorithm. During model construction, special attention should be paid to the Raman characteristic peak regions related to the molecular vibration of each component, including the C-H stretching vibration region of glucose, the C-O stretching vibration region of BDO and ethanol, and the C=O carbonyl stretching vibration region of acetic acid. After the model is built, it must undergo a strict verification procedure, including internal cross-validation and external validation using an independent test data set, to ensure the prediction accuracy and generalization ability of the model and prevent data overfitting.

[0041] Metabolic stress index model parameterization: Using the verified chemometrics model, the full-process spectral data of all historical batches are converted into continuous, high-resolution concentration curves, and combined with the final fermentation performance evaluation index of each batch, the optimal weighting coefficient used in the calculation of the metabolic stress index is calculated and determined through the aforementioned multiple regression or machine learning method.

[0042] Precise amount of control threshold: Select the top 10% of the "golden batch" with the best performance, and conduct detailed statistical analysis on the whole process of metabolic stress index curve. The mean value of these curves plus or minus two standard deviations is defined as the low stress optimal interval in the control target. The 95th percentile value of these curves is set as the early warning threshold for triggering correction control.

[0043] System initialization deployment: Load the final determined chemometrics model file, metabolic stress index weighting coefficient, and quantitative control threshold parameter into the central processor or industrial computer of the fermentation control system to complete the system deployment and preparation work.

[0044] Phase two: online dynamic optimization; This phase is the real-time closed-loop control process performed during each actual fermentation production.

[0045] Start and high-frequency monitoring: Start a new BDO fermentation process, and the system automatically starts the online Raman spectrometer, uses a 785 nm laser wavelength to reduce biological fluorescence interference, sets the acquisition parameters to ensure data quality and real-time performance.

[0046] Data real-time processing cycle: Spectrum input: Collect a new frame of raw Raman spectrum data.

[0047] Spectrum pretreatment: Automatically perform a series of pretreatment steps, first subtract the dark spectrum, then use the median filter algorithm to remove cosmic ray-induced peak noise, then apply the asymmetric least squares method for baseline correction to eliminate fluorescence background, and finally normalize the spectrum to correct for small fluctuations in laser power.

[0048] Concentration analysis: Input the clean spectrum data into the loaded chemometrics model to instantly calculate the current concentrations of BDO, glucose, acetic acid, ethanol, etc.

[0049] Rate calculation: Add the newly analyzed concentration points to their respective time series, and the system uses the Savitzky-Golay filter to process the time series window containing the latest data points. This filter can directly calculate the first derivative of the smoothed curve, i.e. the instantaneous generation rate of each component, while smoothing the noise through local polynomial fitting.

[0050] State quantification: Call the metabolic stress index model parameters, and substitute the instantaneous generation rate of each component into the weighted operation to calculate the metabolic stress index value at the current time, and at the same time calculate its linear regression slope in the recent time window as its change gradient.

[0051] State determination and decision: The system compares the calculated metabolic stress index and its gradient with the preset threshold to automatically determine the state level to which the current process belongs.

[0052] Control instruction output: According to the determination result, the control system sends specific adjustment instructions to the programmable logic controller or distributed control system at the bottom layer of the fermenter through the standard industrial communication protocol OPCUA.

[0053] Uninterrupted cycle execution: The system returns to the input step of the real-time data processing cycle and continues the closed-loop operation of "monitoring-treatment-decision-execution" until the operator issues the fermentation end instruction. Specific embodiment three: As Figures 1-2 shown, in a Bi-BDO fermentation pH dissolved oxygen dynamic optimization method based on online Raman spectrum, the system determination scheme is: The system has a three-level state determination scheme that automatically switches according to the real-time quantitative results of the metabolic stress index, achieving automation and standardization of decision-making.

[0055] State level State name Decision condition System action Level 1 Steady-state optimization zone Metabolic stress index value is within the low-stress optimal interval, and its linear regression slope over the past 15 minutes is less than a pre-set positive micro-threshold. Maintain: No active intervention, keep current pH and DO setpoints. System continues high-frequency monitoring. Level 2 Metabolic warning zone Metabolic stress index value exceeds the "low-stress optimal interval" but is below the "warning threshold", or its linear regression slope over the past 15 minutes is consistently positive and statistically significant. Feedforward fine-tuning: Decision as a precursor to metabolic imbalance. Initiate a minimal disturbance regulation sequence, preferentially fine-tuning the DO setpoint, and closely observe the subsequent metabolic stress index response. Level 3 Imbalance control zone Metabolic stress index value has clearly broken the "warning threshold". Correction control: Decision as a clear shift in metabolism has occurred. Execute a more forceful co-regulation, simultaneously adjusting both pH and DO setpoints according to a pre-set control map. System control logic: Steady-state logic: In level 1 state, the control system is in "observer" mode, and its core task is to ensure the continuity and accuracy of monitoring data, providing high-quality basis for state change.

[0056] Feedforward logic: In level 2 state, the system enters "precautioner" mode, and the core of this logic is an active "detection-evaluation-decision" sequence. The system first increases the DO set value by a small step, 0.5% saturation, and then continuously calculates the response gradient of the metabolic stress index within the next 20-minute observation window. When the gradient turns negative, it proves that the adjustment is effective, and the system will lock the new DO set value and try to return to level 1. When the gradient is still positive or has no significant change, the system determines that simply adjusting the dissolved oxygen is ineffective, and will restore the original DO value and try to slightly decrease the pH set value by 0.05 units, and enter the observation and evaluation cycle again. This logic aims to quickly locate and solve potential problems with minimal environmental disturbance.

[0057] Correction logic: in level 3 state, the system enters "intervener" mode, where no tentative fine-tuning is performed, but a multi-dimensional control map built in offline modeling stage is directly called. This map is a response surface model, whose input variables include current metabolic stress index value, main by-product relative contribution to index exceeding, and current substrate concentration, and its output is the calculated adjustment of pH and DO setpoint, i.e. ΔpH and ΔDO. The system adds current setpoint to these two adjustments to get new control target, in order to pull metabolic state back to pre-warning zone as fast as possible.

[0058] Data content processing, input and output: System input: Real-time dynamic data: raw spectral data collected by Raman spectrometer, represented as high-dimensional float number vector containing photon counting value and Raman shift value, real-time pH and DO measurement value from fermenter online sensor, as control feedback and current state record.

[0059] Static configuration data: chemometrics model file, containing regression coefficient vector for prediction, metabolic stress index model parameters, including by-product Raman characteristic peak list for monitoring and corresponding weighting coefficient value, control scheme parameters, including upper and lower limit of "low stress optimal interval" and specific value of "pre-warning threshold", multi-dimensional control map model file used by correction logic.

[0060] Core data processing: Spectrum pretreatment: automatically execute standardization process including dark spectrum subtraction, cosmic ray removal, asymmetric least squares baseline correction and spectrum total area normalization.

[0061] Concentration prediction: perform matrix operation between pretreated spectrum vector and regression coefficient vector of chemometrics model, output concentration prediction value of each component.

[0062] Time series processing and rate calculation: use a Savitzky-Golay filter containing 11 data point window to perform smoothing and first-order differential processing on concentration time series, directly output instantaneous generation rate.

[0063] Metabolic stress index synthesis: perform linear combination operation on instantaneous generation rate of each component according to preset weighting coefficient, synthesize single metabolic stress index scalar value.

[0064] State determination and decision: input synthesized index and its gradient into preset determination rule and control logic module for comparison and judgment.

[0065] System output: Control instruction: send new pH setpoint to the controller of the pH control unit, send new DO setpoint to the controller of the DO control unit. All control instructions are transmitted safely and reliably through industrial standard protocols such as OPCUA.

[0066] Process data and logs: record and store the concentration curves of each component, the change curves of metabolic stress index in real time in the form of CSV files or writing into time series databases, and record in detail each time the system state is switched in the dedicated event log, as well as the specific values of all pH and DO regulation actions performed by the system automatically, the execution time and response results. These structured data provide a solid foundation for subsequent process analysis, fault diagnosis and continuous iterative optimization of the model. Specific embodiment four: As Figures 1-2 shown, according to the content in the above specific embodiments, the following is further disclosed, and the following provides specific use cases: Background and scenario: In a 500-liter Bi-BDO fermentation tank of pilot scale, the fermentation process enters the late logarithmic growth phase, which is the key stage of the fastest BDO synthesis rate. The fermentation tank uses a standard distributed control system, and the pH and DO are set to the best constant values in historical experience, which are 6.8 and 15% saturation, respectively. During this period, due to the slight fluctuation of the total air pipe pressure in the workshop, the air flow supplied to the fermentation tank decreases slightly, but the decrease is very small, and the reading of the DO online probe still fluctuates between 14.8% and 15% in a narrow range, and does not trigger the traditional DO low alarm.

[0068] System monitoring and data performance: At the 28th hour of fermentation, the operator sees from the monitoring interface that the conventional parameters such as pH, DO, temperature and feed rate are all normal, and the slope of the BDO concentration curve also remains good, however, the background data of the dynamic optimization system of the present application shows abnormal signals.

[0069] From 28 hours and 10 minutes, the metabolic stress index calculated by the system starts to slowly but continuously climb from the steady-state baseline of-0.5. At 28 hours and 40 minutes, the index has reached +1.2, and by analyzing its composition, the system finds that the growth of the index is mainly due to the significant increase in the instantaneous generation rate of ethanol, at the same time, the system calculates the linear regression slope of the index in the past 30 minutes as a positive value with high confidence, indicating that it is a continuous deterioration trend, rather than random data fluctuation.

[0070] System determination and control execution: At 28h41min, the system automatically switches the current state from the steady-state optimization zone of level 1 to the metabolic warning zone of level 2 according to the decision scheme, the system determines that although the apparent parameters are normal, a slight redox imbalance has occurred inside the cell, i.e. a reduction pressure is generated, causing part of the carbon flow to divert to ethanol synthesis to regenerate NAD+.

[0071] The system immediately starts the feedforward fine-tuning control logic, first, it issues an instruction to the DO control unit to slightly increase the dissolved oxygen setpoint from 15.0% to 15.5%, in the next 20 minutes, the system continuously monitors the response of the metabolic stress index, the data shows that after the DO is increased, the instantaneous generation rate of ethanol begins to decrease, and the metabolic stress index begins to fall from the high point of +1.2, to 29h15min, the index has successfully returned to -0.2, re-entering the low stress optimal interval.

[0072] Results and values: Thanks to the proactive intervention of the system, a "silent" metabolic drift triggered by a slight oxygen deficiency, which cannot be detected by traditional methods, is successfully nipped in the bud, without the system, this state would continue to deteriorate, and the operator would not discover the accumulation of ethanol and the decrease in BDO yield from offline HPLC data for several hours, but by then irreversible carbon source waste has already been caused, this case fully demonstrates the ability of the method to prevent a disaster, by making minimal intervention at the initial stage of the problem, it guarantees the stability of the metabolic flow in the key production stage, and estimates that it has improved the final BDO yield of this batch by at least 5%. Specific embodiment five:

[0074] As Figures 1-2 shown, according to the content in the above specific embodiments, the following content is further disclosed, the following provides a specific use case: Background and scenario: In an industrial production level 50 cubic meter BDO fermenter, the fermentation process is in the stable period, due to the failure of the pump in the downstream processing unit, the back pressure of the exhaust gas pipeline of the fermenter instantaneously increases, seriously affecting the mass transfer efficiency of the fermenter, causing the dissolved oxygen in the tank to rapidly drop from 20% to 5% in a short time, although the engineer repaired the fault in 15 minutes and manually restored the dissolved oxygen to the set value of 20%, but this strong anoxic disturbance has caused a huge metabolic impact on the strain.

[0075] System monitoring and data performance: During the sudden drop of DO, the metabolic stress index of the system soared from -0.3 to +8.5 in just 15 minutes, far exceeding the preset warning threshold of +3.0. The analysis of the contribution of the index showed that the instantaneous generation rates of both acetic acid and ethanol increased explosively, indicating that the energy metabolism and redox balance of the cells collapsed simultaneously under the oxygen stress.

[0076] When the engineer manually restored the DO to 20%, the conventional control system completed its task, but the system of the present application showed that, despite the restoration of DO, the metabolic stress index only decreased slightly to +7.9 in a short time, still at a very high stress level, indicating that the metabolic network of the cells had fallen into a "vicious cycle" and could not recover on its own by simply restoring the DO.

[0077] System determination and control execution: Upon the sudden drop of DO, the system immediately determined that the state was in the unbalanced control zone of level 3. After the DO was manually restored, the system determination process was still at level 3, and the correction control logic was immediately started.

[0078] Instead of making exploratory fine-tuning, the system directly called the built-in multi-dimensional control atlas. According to the current input of "MSI value = +7.9, acetic acid and ethanol jointly dominant, substrate concentration sufficient", the optimal correction strategy calculated by the atlas model was: first, lower the pH set value from 6.5 to 6.2 to suppress the dissociative toxicity of acetic acid and change the metabolic balance; second, further increase the DO set value from 20% to 25% for a short time of "oxygen overexposure" to help the cells quickly consume the accumulated reducing power.

[0079] The system automatically executed this coordinated regulation instruction.

[0080] Results and values: Within about 2 hours after the system executed the coordinated regulation, the metabolic stress index was quickly and effectively suppressed, successfully decreasing from +7.9 to +2.5, entering the warning zone. Then the system automatically switched back to the feedforward fine-tuning logic, gradually guiding the pH and DO back to the regular set values within the next 4 hours, and finally returning the metabolic stress index to the optimal state. Although this strong disturbance caused some loss in the production stage, through the precise, rapid, and multi-variable coordinated correction control of the system, the strain was successfully "rescued" from severe metabolic collapse, avoiding the disastrous consequences of discarding the entire batch of fermentation broth and recovering significant economic losses.

[0081] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it is intended to be limited only by the words recited in the appended claims. It is to be understood that the terms "including", "comprising", "consisting" and variations thereof do not preclude the addition of further integers to the combinations of integers specified in the claims. It is to be understood that the phraseology and terminology employed herein are for the purpose of description and illustration only and that the use of particular terms to describe particular embodiments is by way of illustration only and not by way of limitation. It is intended that the description of the various embodiments will enable anyone skilled in the art to practice the application as claimed. Description of the various embodiments is intended to cover any and all modifications and variations of the various embodiments including combinations of features of the various embodiments. It is intended that the scope of the application disclosed herein should not be limited by the particular forms set forth in the description above but should include all embodiments which do not depart from the spirit and scope of the present application. It is further intended that the scope of the claims should not be limited to the particular forms set forth in the description above but should include all embodiments that do not depart from the spirit and scope of the present application.

[0082] While the embodiments of the application have been shown and described herein, it is to be understood that the application is not limited to these embodiments. Rather, numerous modifications can be made by those skilled in the art without departing from the spirit and scope of the present application, which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation based on online Raman spectroscopy, characterized in that: Comprising the following steps: Sp1. Online spectrum and concentration data acquisition: during the BDO fermentation process, the online Raman spectrum data of the fermentation broth is continuously collected, and the concentration time series data of the target product BDO and the by-products related to cell metabolic stress are obtained by real-time analysis of the chemometrics model; Sp2. Metabolic stress state quantification: based on the concentration time series data obtained in Sp1, the first derivative is taken to obtain the instantaneous generation rate of the target product and the by-products, and a comprehensive index for real-time quantification of the overall metabolic stress level of the fermentation system is calculated according to the preset function relationship, defined as the metabolic stress index; Sp3. Metabolic imbalance trend prediction: analyze the time series of the metabolic stress index calculated in Sp2 to obtain its current value and its time change gradient, when the value of the metabolic stress index exceeds the preset baseline threshold, or its time change gradient presents a continuous positive growth trend, it is determined that there is a precursor of metabolic imbalance in the fermentation process; Sp4. Adjustment decision and execution: when Sp3 determines that there is a precursor of metabolic imbalance, the preset adjustment logic is started to apply a small amplitude adjustment to the pH and dissolved oxygen control variables, and returns to Sp1 to continuously monitor the inhibition effect of the adjustment on the metabolic stress index; Sp5. Closed-loop optimization: according to the actual inhibition effect of the adjustment amount applied in Sp4 on the metabolic stress index, the direction and amplitude of subsequent adjustment are adaptively adjusted, and the metabolic stress index is optimized and maintained within the preset optimal interval by dynamically adjusting the pH and dissolved oxygen, so as to realize the dynamic optimization of the fermentation process.

2. The method for dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation based on online Raman spectroscopy according to claim 1, characterized in that: The by-products related to cell metabolic stress in Sp1 are one or a combination of acetic acid, ethanol, lactic acid or pyruvic acid.

3. The method of claim 1, wherein the method is a method for dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation based on online Raman spectroscopy. In Sp2, the weighting coefficients for weighting the instantaneous generation rates of the by-products and the target product BDO to calculate the metabolic stress index are obtained by statistical analysis and optimization calculation of historical fermentation batch data with different pH and dissolved oxygen conditions, and the optimization target is to maximize the negative correlation between the metabolic stress index and the final BDO yield.

4. The method for dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation based on online Raman spectroscopy according to claim 1, characterized in that: The baseline threshold of the metabolic stress index in Sp3 and the optimal interval are determined by statistical analysis of the metabolic stress index data of high-yield stable fermentation batches.

5. The method of claim 1, wherein the method is a method for dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation based on online Raman spectroscopy. The adjustment logic in Sp4 is a hierarchical response logic: when it is determined that there is a slight imbalance precursor, only the dissolved oxygen control variable is fine-tuned; when it is determined that there is a significant imbalance precursor, the dissolved oxygen and pH control variables are adjusted coordinately.

6. The method of claim 1, wherein the method is a method of online Raman spectroscopy-based dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation. The adjustment amount in Sp4 is a small amplitude adjustment amount of 0.1% to 5% of the current set value.

7. The method of claim 1, wherein the method is a method of online Raman spectroscopy-based dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation. The closed-loop optimization in Sp5 is an adaptive feedback optimization, which uses hill climbing algorithm or particle swarm optimization algorithm to minimize the metabolic stress index as the target to search for the optimal pH and dissolved oxygen adjustment amount online.

8. The system for online Raman spectroscopy-based dynamic optimization of pH and dissolved oxygen in a Bi-BDO fermentation according to any one of claims 1-7, characterized in that, Comprising: Bi-BDO bioreactor equipped with pH and dissolved oxygen sensing and execution mechanism; Online Raman spectrum analyzer; Central processing unit configured to: receive and analyze Raman spectrum data to obtain concentration time series; According to Sp2, the metabolic stress index is calculated in real time; According to Sp3, the metabolic imbalance trend is predicted in real time; According to Sp4 and Sp5, adjustment instructions are generated and sent to the actuators of the fermenter to form a closed-loop control.

9. The system of online Raman spectrum-based dynamic optimization of pH and dissolved oxygen for a Bi-BDO fermentation according to claim 8, characterized in that: The central processor further comprises a historical database and a model optimization module, which are used to store historical fermentation data and to perform offline optimization and update of the weighting coefficients used in the calculation of the metabolic stress index.

Images