Method and system for adjusting pH value in microbial fermentation process

Abstract

The invention belongs to the technical field of microbial fermentation engineering, and provides a method and a system for adjusting a pH value in a microbial fermentation process, and the method comprises the following steps: screening a pH jump batch and a buffer capacity sudden drop batch based on historical data, verifying a buffer saturation causal relationship, and carrying out piecewise linear regression on the buffer capacity and pH of the buffer saturation batch to obtain the pH value of the buffer saturation batch. A system-level buffer saturation point is determined, then the buffer capacity trend is monitored in real time, the saturation moment is predicted according to sequence characteristics, finally, the acid / alkali regulation amount is reversely deduced based on a historical correlation model, the saturation point and a fermentation end point are synchronized, and the problems of invalid supplementation-sudden rising and sudden falling caused by biological buffer system enhancement in high-density fermentation are solved. The pH regulation and control is converted from lagging response to pre-judgment in advance, the yield and the stability of fermentation products are remarkably improved, and the method is suitable for high-density fermentation scenes such as escherichia coli and saccharomycetes.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of microbial fermentation engineering, and particularly relates to a method and system for adjusting pH value in a microbial fermentation process. BACKGROUND

[0002] In a microbial fermentation process, pH value is a core environmental parameter affecting cell growth, metabolism and product synthesis, and its stability directly determines fermentation efficiency and product quality. With the upgrading of fermentation technology to "high density and high product concentration", the buffering characteristics of the fermentation broth have changed significantly: in high-density fermentation, the cell concentration can reach 10-20 g / L (dry weight), and the cell itself carries a large number of amino, carboxyl and other dissociable groups in macromolecules such as proteins and nucleic acids, forming a "biological buffer system" and increasing the buffering capacity of the fermentation broth from 0.05-0.1 mol / (L·pH) in conventional fermentation to 0.2-0.3 mol / (L·pH).

[0003] The existing pH adjustment strategy does not adapt to the above-mentioned changes in characteristics, resulting in prominent problems of regulation failure: on the one hand, due to the enhancement of the biological buffer system, when the acid / alkali is added according to the conventional dosage, there is no obvious change in pH (the addition is ineffective), and the operator is forced to increase the addition amount; on the other hand, when the biological buffer system and the chemical buffer reach the "saturation point", a small amount of acid or alkali addition will cause a sharp rise or fall in pH, thereby inhibiting cell metabolism and reducing product synthesis efficiency. The existing technology has three major defects: causality recognition is missing: it is difficult to distinguish whether the pH jump is caused by buffer saturation, contamination or metabolic mutation, which is easy to misjudge the control direction; there is no quantitative standard for the saturation point: the saturation of the buffer is determined by the experience of the operator, and there is no objective definition of the saturation point based on data; the control is lagging: the adjustment is made only after the pH appears to be abnormal, and the dynamic trend of the buffer capacity is not considered in advance; Therefore, the present application provides a method and system for adjusting pH value in a microbial fermentation process. SUMMARY

[0004] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background.

[0005] The technical scheme adopted by the present application to solve its technical problems is: a method for adjusting pH value in a microbial fermentation process, comprising: Step 1: performing jump analysis on the pH values before and after acid / alkali addition in the microbial fermentation process of different batches of the same fermentation system, dividing the historical batches into jump batches and non-jump batches, identifying the buffer capacity jump batches through buffer capacity verification, and performing overlap analysis on the jump batches, non-jump batches and buffer capacity jump batches, respectively, and determining whether the pH jump before and after acid / alkali addition is caused by buffer saturation; Step two: If yes, segment fitting is performed on the acid / base make-up amount and pH of the buffer saturation batch to determine the buffer saturation point, and stability analysis is performed on the buffer saturation points of each buffer saturation batch to determine the buffer saturation point of the current fermentation system; Step three: Real-time monitoring and analysis are performed on the buffer capacity trend of the current fermentation process, and the buffer saturation time is predicted in combination with the buffer saturation point of the current fermentation system; Step four: The predicted buffer saturation time is compared with the fermentation end time, if the predicted saturation time is before the fermentation end time, the acid / base addition amount and buffer capacity between the critical time of the buffer saturation batch and the fermentation end time are analyzed for correlation, an association model is obtained, and the target buffer capacity consumption rate is determined according to the current buffer capacity, which is substituted into the association model to obtain the acid / base control amount.

[0006] Further, the division method of the jump batch and the non-jump batch is: The time point of each addition of acid / base and the corresponding pH value in the microbial fermentation process are obtained, and the pH change amount before and after the addition of acid / base is calculated, which is integrated into a pH change amount time sequence in chronological order; Based on the historical data of normal batches of the fermentation system, the mean and standard deviation of the pH change amount in the normal fermentation stage are calculated, and the jump threshold is set based on the 3σ principle, wherein the normal batch is a historical batch with stable pH value in the fermentation process, which is determined by calculating the standard deviation of the pH value; If the pH change amount is greater than the jump threshold, the addition of acid / base is marked as pH jump addition; In the pH change amount sequence, the pH change amount is compared with the invalid threshold from the pH jump addition to the front, if the pH change amount is less than the invalid threshold, the addition of acid / base is marked as invalid addition; If the invalid addition is followed by pH jump addition, the historical batch is marked as a jump batch, and the pH jump time point is recorded; On the contrary, if there is any other situation except the invalid addition followed by pH jump addition, it is a non-jump batch.

[0007] Further, the identification method of the buffer capacity jump batch is: According to the calculation formula of the buffer capacity, the acid / base concentration change amount and the pH change amount before and after each addition are proportionally calculated to obtain the buffer capacity, and the buffer capacity time sequence is integrated according to the addition sequence; Based on the normal batch data of the same fermentation system, the normal buffer capacity interval is determined, and the buffer capacity in the buffer capacity time sequence is compared with the normal buffer capacity interval; If the buffer capacity time series all meet the sudden drop determination conditions, the historical batch is marked as a buffer capacity sudden drop batch, and the time point when the buffer capacity first falls below the normal buffer capacity interval, i.e. the time point when the buffer capacity starts to decrease, is recorded; wherein the sudden drop determination conditions include: determination condition one: there is any and more buffer capacity in the buffer capacity time series below the normal buffer capacity interval; determination condition two: the buffer capacity time series presents a continuous decrease; determination condition three: the buffer capacity value after the buffer capacity starts to decrease is integrated into a sequence, and according to the mathematical definition of limit, it is judged that the limit of the sequence is 0.

[0008] Further, the judgment method of whether the pH value jump before and after the acid / alkali addition is caused by buffer saturation is: For any one jump batch, if the jump batch is a buffer capacity sudden drop batch, the jump batch is marked as a jump overlap batch; For any one non-jump batch, if the non-jump batch is a buffer capacity sudden drop batch, the non-jump batch is marked as a non-jump overlap batch; For any one jump overlap batch, the deviation of the pH jump time point and the buffer capacity sudden drop time point is calculated, and compared with the preset deviation, if the deviation is less than the preset deviation, and the pH jump time point is after the buffer capacity sudden drop time point, the jump overlap batch is marked as a buffer saturation batch; The proportion of buffer saturation batches in jump batches is counted to obtain the buffer saturation proportion; The proportion of non-jump overlap batches in non-jump batches is counted to obtain the non-jump overlap proportion; The buffer saturation proportion and the non-jump overlap proportion are compared with the preset proportion respectively, if the buffer saturation proportion is greater than the preset proportion, and the non-jump overlap proportion is less than the preset proportion, the jump is caused by buffer saturation.

[0009] Further, the determination process of the buffer saturation point of the current fermentation system is: The buffer capacity and pH data of the buffer saturation batch are extracted, and for any one buffer saturation batch: A quadratic polynomial fitting is performed with the pH value as the independent variable and the corresponding buffer capacity as the dependent variable, and the fitting parameters are solved by the least square method to obtain the fitting curve; A sliding window is set, and the fitting curve is traversed by the sliding window for point-by-point sliding; For each sliding window, the local slope in the sliding window is calculated by linear regression, and the slope change rate of adjacent sliding windows is calculated; The rate of change of the slope of adjacent sliding windows is compared with a preset rate of change, and if the rate of change of the slope of adjacent sliding windows is greater than the preset rate of change, the junction point of the adjacent sliding windows is taken as an initial segmentation point; Based on the initial segmentation point, the data is divided into two segments, and linear fitting is performed on each segment to obtain two linear equations. A sliding window is set, the linear regression slope in each window is calculated, and the rate of change of the slope of adjacent windows is calculated. The rate of change of the slope of adjacent windows is compared with a preset rate of change, and if the rate of change of the slope of adjacent windows is greater than the preset rate of change, the junction point of the adjacent windows is taken as an initial segmentation point; Based on the initial segmentation point, the data is divided into two segments, and linear fitting is performed on each segment to obtain two linear equations. F test is performed on the two fitting equations. The ratio of the slopes of the two fitting equations is calculated and compared with a preset threshold. If the ratio of the slopes is greater than the preset threshold, and the F test determines that there is a statistical difference between the two slopes, then the intersection point of the two straight lines is the buffer saturation point under the current fermentation system.

[0010] Further, the prediction process of the buffer saturation time is: The current fermentation process is monitored in real time, and the buffer capacity is calculated in real time after each monitoring. A buffer capacity threshold is set, and after each monitoring, the real-time buffer capacity is compared with the buffer capacity threshold. If the current buffer capacity exceeds the buffer capacity threshold, calculate the difference between the current buffer capacity and the buffer saturation point to obtain an approximation value. If the approximation value is less than or equal to a preset threshold, mark the time when the buffer capacity reaches the buffer capacity threshold as a critical time. The time period between the critical time and the current time is marked as a buffer warning window period, and the buffer capacity monitoring values in the buffer warning window period are integrated into a buffer warning capacity monitoring sequence. Linear fitting is performed on the buffer warning capacity monitoring sequence, the least squares method is used to solve the fitting equation, the goodness of fit of the fitting equation is calculated, and if the goodness of fit is greater than or equal to an empirical threshold, the buffer warning capacity monitoring sequence is linear, otherwise it is nonlinear. If the buffer warning capacity monitoring sequence is linear, the time when the fitting equation is equal to the buffer saturation point is the predicted saturation time. If the buffer warning capacity monitoring sequence is nonlinear, a nonlinear prediction model is constructed based on the real-time monitoring index time series data of the fermentation process to obtain the predicted saturation time.

[0011] Further, the way to construct a nonlinear prediction model to predict the buffer saturation time is: The real-time monitoring index time series data of the fermentation process includes: buffer warning capacity monitoring sequence, real-time pH value, acid / base cumulative addition amount, buffer concentration, cell concentration, and metabolic product concentration. All monitoring indicators are aligned according to timestamps and standardized; Historical information is extracted through a sliding window to construct input-output sample pairs; A two-way LSTM+Dropout+fully connected layer structure is adopted to capture nonlinear time series features in layers and construct an LSTM model; Real-time monitoring data in the buffer warning window period are generated as input samples according to a sliding window, features are automatically calculated and standardized, the input samples are input into the LSTM model, and the remaining saturation time is output; The current time and the remaining saturation time are summed to obtain the buffer saturation prediction time.

[0012] Further, the process of correlating the acid / base addition amount and the buffer capacity at the buffer saturation batch critical time and the fermentation end time is: The acid / base cumulative addition amount of the fermentation process and the buffer capacity at the corresponding time are normalized according to the relative time, and the time is converted into the relative length of time from the saturation time, taking the buffer saturation time of the buffer saturation batch as the reference; The addition amount of acid / base and the buffer capacity before and after each addition of acid / base in the buffer warning window period of the buffer saturation batch are extracted; For the buffer capacity of any buffer saturation batch, the time differential is performed to obtain the instantaneous consumption rate: wherein, is the monitoring interval, and the negative sign indicates that the buffer capacity decreases over time; The instantaneous addition rate of acid / base is used as the independent variable, and the buffer capacity consumption rate is used as the dependent variable, and a regression equation is fitted by using the least squares method.

[0013] Further, the determination method of the target buffer capacity consumption rate and the acid / base control amount is: The residual buffer capacity is obtained by subtracting the buffer capacity at the buffer saturation point from the buffer capacity at the current time; The time period between the current time and the fermentation end time is marked as the allowed consumption time; The target buffer capacity consumption rate is obtained by proportionally calculating the residual buffer capacity and the allowed consumption time; The allowed acid / base addition rate and cumulative addition amount are obtained by substituting the target buffer capacity consumption rate into the correlation model.

[0014] A pH value regulation system for a microbial fermentation process includes the following modules: The buffer saturation judgment module: the pH values before and after the acid / alkali supplement in the microbial fermentation process of the historical different batches under the same fermentation system are analyzed by jump analysis, the historical batches are divided into jump batches and non-jump batches, and the buffer capacity verification is carried out to identify the buffer capacity sudden change batches, the jump batches, the non-jump batches and the buffer capacity sudden change batches are analyzed by overlap analysis respectively, and whether the pH value jump before and after the acid / alkali supplement is caused by buffer saturation is judged; The buffer saturation point determination module: if yes, the acid / alkali supplement amount and the pH of the buffer saturation batch are segmented and fitted to determine the buffer saturation point, and the stability of the buffer saturation point of each buffer saturation batch is analyzed to determine the buffer saturation point of the current fermentation system; The buffer saturation time prediction module: the buffer capacity trend of the current fermentation process is monitored and analyzed in real time, and the buffer saturation time is predicted in combination with the buffer saturation point of the current fermentation system; The acid / alkali control amount calculation module: the predicted buffer saturation time is compared with the fermentation end time, if the predicted saturation time is before the fermentation end time, the acid / alkali addition amount between the critical time of the buffer saturation batch and the fermentation end time is associated with the buffer capacity to obtain an association model, and the target buffer capacity consumption rate is determined according to the buffer capacity at the current time, and the acid / alkali control amount is obtained by substituting the acid / alkali control amount into the association model.

[0015] The beneficial effects of the present application are as follows: through jump analysis and buffer capacity verification of historical data, the causal recognition accuracy of pH jump is improved, interference factors such as contamination and equipment failure are effectively excluded, blind control is avoided, the causal relationship is accurately identified, segmented linear regression combined with multi-batch stability analysis is adopted, a unified buffer saturation point quantitative standard is provided for the fermentation system for the first time, the subjective defects of traditional methods relying on experience are solved, the buffer saturation point is quantified, based on buffer capacity real-time monitoring and linear / nonlinear fitting model, the saturation time can be predicted in advance, the control window is expanded from post-exception response to pre-exception prevention, the pH fluctuation range is controlled, the advance prediction is realized, the control amount is back calculated through the historical association model, the biological buffer characteristics of high-density fermentation can be dynamically adapted, the problems of invalid supplement and sudden rise and fall are avoided, the product yield is improved, the acid / alkali consumption cost is reduced, the high-density fermentation is adapted, the matching adjustment system realizes data acquisition-analysis-prediction-control closed loop, can be seamlessly connected with the existing fermentation tank DCS system, does not need large-scale equipment modification, is easy to popularize, and has strong system integration. BRIEF DESCRIPTION OF DRAWINGS

[0016] The present application will be further described below with reference to the drawings.

[0017] Figure 1 It is a step flow chart of the pH value adjustment method of the microbial fermentation process described in embodiment 1 of the present application. Figure 2 is a logic judgment chart of a pH value adjustment method of a microbial fermentation process according to the embodiment 1 of the present application; Figure 3 is a program block diagram of a pH value adjustment system of a microbial fermentation process according to the embodiment 2 of the present application. DETAILED DESCRIPTION

[0018] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application is further described below in combination with specific embodiments.

[0019] Embodiment 1: Please refer to Figure 1 The pH value adjustment method of a microbial fermentation process according to the embodiment of the present application includes the following steps: Step 1: The pH value jump before and after acid / alkali addition in the microbial fermentation process of the same fermentation system under historical different batches is analyzed, the historical batches are divided into jump batches and non-jump batches, and the buffer capacity verification is performed to identify the buffer capacity sudden change batches, the overlap analysis is performed on the jump batches, the non-jump batches and the buffer capacity sudden change batches, and it is judged whether the pH value jump before and after acid / alkali addition is caused by buffer saturation or not; In step 1, the fermentation system includes: microbial consistency (same strain), fermentation substrate consistency (same carbon source, nitrogen source, trace element type and concentration), process parameter consistency (same temperature, stirring speed, aeration rate, dissolved oxygen control strategy), and buffer system consistency (same buffer type and initial concentration); The data related to pH adjustment in the historical batches are obtained, including: Basic process data: fermentation time, temperature, stirring speed, aeration rate, dissolved oxygen, and cell concentration; pH related data: real-time pH value, pH set target value, acid / alkali addition type and addition volume / rate; Buffer system data: initial buffer concentration, and buffer addition record in the fermentation process; In step 1, the process of judging whether the pH jump is caused by buffer saturation or not includes: For any historical batch: The time point of each addition of acid / alkali in the microbial fermentation process and the corresponding pH value are obtained, and the pH change amount before and after the addition of acid / alkali is calculated , which is integrated into a pH change amount time sequence in sequence; Based on the historical data of normal batches of the same fermentation system, the mean and standard deviation of the pH change amount in the normal fermentation stage are counted, and the jump threshold is set based on the "3σ" principle; The pH change amount in the pH change amount time sequence is compared with the jump threshold value, and if the pH change amount is greater than the jump threshold value, the acid / alkali addition is marked as a pH jump addition; In the pH change amount sequence, the pH change amount is compared with the invalid threshold value from the pH jump addition, and if the pH change amount is less than the invalid threshold value, the acid / alkali addition is marked as an invalid addition; The number of consecutive invalid additions before the pH jump addition is counted, and if one pH jump addition occurs after one or more consecutive invalid additions, the batch is marked as a jump batch, and the pH jump time point is recorded; On the contrary, it is marked as a non-jump batch; The buffer capacity before and after each acid / alkali addition is calculated, and the formula is wherein, is the total amount of substance of the strong acid or strong base added, and the calculation method is the product of the addition volume and the acid / alkali concentration, V is the actual volume of the fermentation system, is the pH value change before and after the addition of acid / alkali; It can be understood that the buffer capacity directly reflects the ability of the fermentation system to resist pH fluctuation. If β is stable in the normal interval, it means that the buffer capacity of the system is sufficient, and a small amount of acid / alkali can be added to maintain pH stability. If β continues to decrease, it means that the buffer system is gradually ineffective. At this time, even if a small amount of acid / alkali is added, it may also cause a sudden change in pH; The buffer capacity time sequence of the whole batch is calculated according to the calculation formula of the buffer capacity; Based on the normal batch data of the fermentation system, the normal buffer capacity interval is determined, and the buffer capacity in the buffer capacity time sequence is compared with the normal buffer capacity interval; If the buffer capacity is lower than the normal buffer capacity interval and continues to decrease to approach 0, it is marked as a buffer capacity sudden drop batch, and the time point when the buffer capacity starts to decrease is recorded; Wherein, the method for approaching 0 is: the buffer capacity value after the buffer capacity starts to decrease is integrated into a sequence, and the limit of the sequence is judged to be 0 according to the mathematical definition of limit; For any jump batch, if the jump batch is a buffer capacity sudden change batch, the jump batch is marked as a jump overlap batch; For any non-jump batch, if the non-jump batch is a buffer capacity sudden change batch, the non-jump batch is marked as a non-jump overlap batch; For any jump overlap batch, the deviation of the pH jump time point and the buffer capacity sudden drop time point is calculated, and compared with the preset deviation. If the deviation is less than the preset deviation, and the pH jump time point is after the buffer capacity sudden drop time point, the jump overlap batch is marked as a buffer saturation batch; Statistically count the proportion of buffer saturation batches in jump batches and the proportion of non-jump overlapping batches in non-jump batches respectively to obtain buffer saturation proportion and non-jump overlapping proportion; Compare the buffer saturation proportion and the non-jump overlapping proportion with the preset proportion respectively, if the buffer saturation proportion is greater than the preset proportion and the non-jump overlapping proportion is less than the preset proportion, then the jump is caused by buffer saturation; The effect of determining whether the jump is caused by buffer saturation is: Through system analysis of historical batch data under the same fermentation system, from the three dimensions of "pH jump identification", "buffer capacity verification" and "two types of feature overlap analysis", finally determine whether the pH jump in the historical batch is caused by buffer system saturation, provide problem root confirmation basis for subsequent regulation: if the jump is not related to buffer saturation, there is no need to optimize the buffer system; if it is related, start subsequent targeted analysis; Step two: if so, segment fitting is performed on the acid / base addition amount and pH of the buffer saturation batch to determine the buffer saturation point, and stability analysis is performed on the buffer saturation points of each buffer saturation batch to determine the buffer saturation point of the current fermentation system; In step two, the determination process of the buffer saturation point includes: Extract the buffer capacity and pH data of the buffer saturation batch, for any one buffer saturation batch: Take the pH value as the independent variable x and the corresponding buffer capacity as the dependent variable y, perform quadratic polynomial fitting, and the fitting equation is y=ax 2 +bx+c, the fitting parameters are solved by least squares method to obtain the fitting curve ; Set a sliding window, and traverse the fitting curve through the sliding window for point-by-point sliding; Among them, the setting of the sliding window can be set according to the speed of change of the buffer capacity with pH, for example: Fast changing system (such as amino acid fermentation, when the pH rises from 6.5 to 7.0, the buffer capacity drops from 4.0 to 0.5 in 10 minutes): small window (2-3 points) is needed to ensure that the steep drop point can be captured (large window will combine the steep drop segment with the flat segment, which will cover the mutation); Slow changing system (such as antibiotic fermentation, the buffer capacity needs more than 1 hour to drop from 3.0 to 0.5): large window (5-6 points) can be used to avoid misjudgment of trend mutation due to random fluctuations (such as single addition error); For each sliding window, calculate the local slope in the sliding window by linear regression, and calculate the slope change rate of adjacent sliding windows; The rate of change of the slope of the adjacent sliding window is compared with a preset rate of change, and if the rate of change of the slope of the adjacent sliding window is greater than the preset rate of change, the junction point of the adjacent sliding window is taken as an initial segmentation point; Based on the initial segmentation point, the data is divided into two segments, and linear fitting is performed respectively to obtain two straight line equations; For any one segment of data, take the pH value as the independent variable x and the buffer capacity as the dependent variable y, and the linear regression equation is y=kx+b, wherein k is the slope and b is the intercept; By least squares method, the fitting equation of the left segment is C=k1x pH+b1, and the fitting equation of the right segment is C=k2x pH+b2; In order to avoid the disconnection of the two straight lines, the buffer capacity C at the segmentation point of the two straight lines is forced to be equal, that is, k1x pH0+b1=k2x pH0+b2; F test is performed on the two fitting equations to determine whether there is a statistically significant difference, specifically: The residual sum of squares of the two fitting equations and the residual sum of squares of the overall fitting are calculated respectively to calculate the F statistic; According to the degrees of freedom df1=p part-p total, df2=n-p total, the critical value Fα of the F distribution table at the significance level α=0.05 is looked up; If F>Fα, there is a statistically significant difference between the two fitting equations; The ratio of the slopes of the two fitting equations, i.e. k2 / k1, is calculated and compared with a preset threshold; If the ratio of the slopes is greater than the preset threshold and the F test determines that there is a statistically significant difference between the two slopes, the intersection point of the two straight lines is the buffer saturation point under the current fermentation system, which is obtained by solving the simultaneous equations; The simultaneous equations are: k1x pH+b1=k2x pH+b2, and the pH value of the buffer saturation point is obtained by solving the simultaneous equations, pH=(b2-b1) / (k1-k2), and the obtained pH value is substituted into any one of the straight line equations to obtain the buffer capacity of the buffer saturation point; The mean of the buffer capacities of the buffer saturation points of all buffer saturation batches is calculated as the buffer saturation point of the current fermentation system; The role of determining the buffer saturation point of the current fermentation system is: By segmenting fitting, statistical testing and multi-batch data integration of the "buffer capacity-pH" data of the buffer saturation batches, the specific parameters (including the corresponding buffer capacity and pH value) of the "buffer saturation point" under the current fermentation system are accurately defined, and the core value is to establish the "buffer system failure critical standard", which provides a clear "early warning benchmark" for subsequent real-time monitoring and prediction; Step three: Real-time monitoring and analysis of the buffer capacity trend of the current fermentation process, and predicting the buffer saturation time based on the buffer saturation point of the current fermentation system; Please refer to Figure 2 As shown in step three, the prediction process of the buffer saturation time includes: Real-time monitoring of the current fermentation process, including real-time pH value, acid / base addition amount, buffer concentration, cell concentration, and metabolite concentration; Real-time buffer capacity is calculated after each monitoring, generating a real-time buffer capacity time series; Set a buffer capacity threshold, and compare the real-time buffer capacity with the buffer capacity threshold after each monitoring; If the current buffer capacity exceeds the buffer capacity threshold, calculate the difference between the current buffer capacity and the buffer saturation point to obtain the proximity value; Compare the proximity value with the preset threshold, if the proximity value is less than or equal to the preset threshold, mark the time when the buffer capacity reaches the buffer capacity threshold as the critical time; Mark the time period between the critical time and the current time as the buffer warning window period, and integrate the buffer capacity monitoring values in the buffer warning window period into a buffer warning capacity monitoring sequence; Linear fitting is performed on the buffer warning capacity monitoring sequence, and the least squares method is used to solve the fitting equation, and the goodness of fit of the fitting equation is calculated, if the goodness of fit is greater than or equal to the empirical threshold, the buffer warning capacity monitoring sequence is linear, otherwise, it is nonlinear; If the buffer warning capacity monitoring sequence is linear, the time when the fitting equation = buffer saturation point is the buffer saturation prediction time; If the buffer warning capacity monitoring sequence is nonlinear, the real-time monitoring index time series data of the fermentation process is used as the core, including: buffer warning capacity monitoring sequence, real-time pH value, acid / base cumulative addition amount, buffer concentration, cell concentration, and metabolite concentration; Align all monitoring indicators according to the timestamp and standardize them; Extract historical information through sliding window and build input-output sample pairs, including: Set the sliding window length L, which is the length of the historical time series of the model input, where the sliding window length is set according to the average length of the buffer warning window period, for example: the average duration of the out-of-bound phase is 30 min, take L = 6 steps, which includes 30 min of historical data; Each input sample includes multivariate time series data within the sliding window, with dimensions (L, M), where L is the sliding window length and M is the total number of monitoring indicators; The output sample is the remaining time from the current time to the buffer saturation time; The structure of bidirectional LSTM+Dropout+fully connected layer is adopted to capture nonlinear time sequence characteristics in layers and construct an LSTM model. The input dimension is specified in the first LSTM layer, and the first layer LSTM outputs a complete time sequence. The second layer LSTM outputs a final integrated feature vector. A Dropout layer is added after the LSTM layer to randomly discard 20% of the neuron connections, avoiding the model from over-relying on local features. The time sequence features extracted by the LSTM are mapped to the predicted value of the remaining saturation time through a fully connected layer. The real-time monitoring data in the buffer warning window period is generated as input samples in a sliding window, the features are automatically calculated and standardized, the input samples are input into the LSTM model, and the remaining saturation time is output. The current time and the remaining saturation time are summed to obtain the predicted saturation time of the buffer; The function of predicting the saturation time is: The static threshold is converted into a dynamic prediction, realizing the transition from post-processing to pre-judgment, and striving for time for timely regulation; Step four: compare the predicted saturation time of the buffer with the fermentation end time, if the predicted saturation time is before the fermentation end time, perform correlation analysis on the acid / base addition amount and the buffer capacity of the buffer saturation batch between the critical time and the fermentation end time, obtain a correlation model, and determine the target buffer capacity consumption rate according to the current buffer capacity, and substitute it into the correlation model to obtain the acid / base control amount; In step four, the process of performing correlation analysis on the acid / base addition amount and the buffer capacity of the buffer saturation batch includes: Compare the predicted saturation time with the fermentation end time; If the predicted saturation time is after the fermentation end time, the buffer system will not be saturated before the fermentation ends, and no regulation is needed; On the contrary, the buffer system will be saturated before the fermentation ends, and regulation is needed to avoid pH out of control affecting the product, specifically: Extract all the key data of the buffer saturation batch, including: Time series data: acid / base cumulative addition amount in the fermentation process, buffer capacity at the corresponding time, fermentation time; Key node data: buffer saturation time of each batch, fermentation end time, acid / base addition amount at the buffer saturation time, acid / base addition amount at the fermentation end time; Normalize the acid / base cumulative addition amount in the fermentation process and the buffer capacity at the corresponding time according to the relative time, take the buffer saturation time of the buffer saturation batch as the reference, convert the time into the relative length of time from the saturation time, and eliminate the influence of the absolute time difference of different batches; The amount of acid / alkali added and the buffer capacity before and after the addition of acid / alkali for each acid / alkali addition in the buffer warning window of the buffer saturation batch; For the buffer capacity of any buffer saturation batch, the time differential is performed to obtain the instantaneous consumption rate: wherein, is the monitoring interval, and the negative sign indicates that the buffer capacity decreases over time; The instantaneous acid / alkali addition rate is used as the independent variable, and the buffer capacity consumption rate is used as the dependent variable. A regression equation is fitted by using the least squares method: If it is a linear relationship: r B =k·r A +b, k is the acid / alkali addition-buffer consumption correlation coefficient, and b is the basic consumption item caused by the acid / alkali produced by the metabolism of the cell; If it is a nonlinear relationship, wherein, a is the quadratic term coefficient, reflecting the accelerated consumption effect of the buffer capacity under high addition rate; The residual buffer capacity is obtained by subtracting the buffer capacity of the buffer saturation point from the buffer capacity at the current time; The time period between the current time and the end of fermentation is marked as the allowed consumption time; The target buffer capacity consumption rate is obtained by proportional calculation of the residual buffer capacity and the allowed consumption time; It should be noted that the target buffer capacity consumption rate should be less than or equal to the average buffer capacity consumption rate between the critical time and the buffer saturation time in the historical batch to avoid excessive restriction of the addition and cause pH out of control; The allowed acid / alkali addition rate and cumulative addition amount are obtained by substituting the target buffer capacity consumption rate into the correlation model; The role of determining the acid / alkali control amount is: Through precise control, the buffer system is prevented from being saturated in advance, pH out of control is prevented, and the fermentation process is ensured to be stable.

[0020] The technical scheme and advantages of the embodiments of the present application are as follows: the pH value jump analysis is performed on the pH values before and after the acid / alkali addition in the microbial fermentation process of historical different batches under the same fermentation system, the historical batches are divided into jump batches and non-jump batches, the buffer capacity verification is performed, the buffer capacity sudden change batches are identified, the overlap analysis is performed on the jump batches, the non-jump batches and the buffer capacity sudden change batches respectively, and it is judged whether the pH value jump before and after the acid / alkali addition is caused by buffer saturation; if yes, the acid / alkali addition amount and the pH of the buffer saturation batches are segmented fitted, the buffer saturation points are determined, the stability of the buffer saturation points of the buffer saturation batches is analyzed, the buffer saturation point of the current fermentation system is determined, the buffer capacity trend of the current fermentation process is monitored and analyzed in real time, the buffer saturation time is predicted in combination with the buffer saturation point of the current fermentation system, the predicted buffer saturation time is compared with the fermentation end time, if the predicted saturation time is before the fermentation end time, the correlation analysis is performed on the acid / alkali addition amount and the buffer capacity between the critical time of the buffer saturation batch and the fermentation end time, the correlation model is obtained, the target buffer capacity consumption rate is determined according to the buffer capacity of the current time, the acid / alkali control amount is obtained by substituting the correlation model. The present application screens the pH jump batches and the buffer capacity sudden change batches based on historical data, verifies the buffer saturation causality, determines the system-level buffer saturation point by performing segmented linear regression on the buffer capacity and the pH of the buffer saturation batches, real-time monitors the buffer capacity trend and predicts the saturation time according to the sequence characteristics, and finally inversely calculates the acid / alkali control amount based on the historical correlation model, synchronizes the saturation point and the fermentation end point, solves the problem of invalid addition-sudden rise-sudden drop caused by the enhancement of the biological buffer system in high-density fermentation, realizes the change from lag response to early prediction in pH regulation, significantly improves the yield and stability of the fermentation product, and is suitable for high-density fermentation scenes such as Escherichia coli and yeast.

[0021] Embodiment 2: please refer to Figure 3 The pH value regulation system for the microbial fermentation process according to the embodiments of the present application includes the following modules: The buffer saturation judgment module: the pH value jump analysis is performed on the pH values before and after the acid / alkali addition in the microbial fermentation process of historical different batches under the same fermentation system, the historical batches are divided into jump batches and non-jump batches, the buffer capacity verification is performed, the buffer capacity sudden change batches are identified, the overlap analysis is performed on the jump batches, the non-jump batches and the buffer capacity sudden change batches respectively, and it is judged whether the pH value jump before and after the acid / alkali addition is caused by buffer saturation; The fermentation system includes: microbial consistency (same strain), fermentation substrate consistency (same carbon source, nitrogen source, trace element type and concentration), process parameter consistency (same temperature, stirring rate, aeration rate, dissolved oxygen control strategy), buffer system consistency (same buffer type and initial concentration); acquire the data related to pH adjustment in the historical batch, including: basic process data: fermentation time, temperature, stirring speed, aeration rate, dissolved oxygen, cell concentration; pH related data: real-time pH value, pH set target value, acid / base addition type and addition volume / rate; buffer system data: initial buffer concentration, buffer addition record during fermentation; The process of determining whether the pH jump is caused by buffer saturation includes: For any one historical batch: acquire the time point of each acid / base addition and the corresponding pH value during microbial fermentation, and calculate the pH change before and after the addition of acid / base , integrate into a pH change time sequence in chronological order; Based on the normal batch historical data of the fermentation system, the mean and standard deviation of the pH change in the normal fermentation stage are calculated, and the jump threshold is set based on the "3σ" principle; Compare the pH change in the pH change time sequence with the jump threshold, if the pH change is greater than the jump threshold, mark the acid / base addition as pH jump addition; In the pH change sequence, traverse from the pH jump addition to the front, compare the pH change with the invalid threshold, if the pH change is less than the invalid threshold, mark the acid / base addition as invalid addition; Statistical number of consecutive invalid additions before pH jump addition, if there is one or more consecutive invalid additions followed by one pH jump addition, mark the batch as jump batch and record the pH jump time point; Otherwise, it is marked as a non-jump batch; Calculate the buffer capacity before and after each addition of acid / base, the formula is , wherein is the total amount of substance of the strong acid or strong base added, calculated as the product of the addition volume and the acid / base concentration, V is the actual volume of the fermentation system, is the change of pH value before and after the addition of acid / base; It can be understood that the buffer capacity directly reflects the ability of the fermentation system to resist pH fluctuations, if β is stable in the normal range, it means that the buffer capacity of the system is sufficient, and a small amount of acid / base can be added to maintain pH stability, if β continues to decrease, it means that the buffer system gradually fails, at this time even a small amount of acid / base may cause pH jump; Calculate the buffer capacity time sequence of the whole process of the batch according to the formula of buffer capacity; Based on the normal batch data of the fermentation system, the normal buffer capacity interval is determined, and the buffer capacity in the buffer capacity time sequence is compared with the normal buffer capacity interval; If the buffer capacity is lower than the normal buffer capacity interval and continues to decrease to approach 0, it is marked as a buffer capacity sudden drop batch, and the time point at which the buffer capacity starts to decrease is recorded; Wherein, the method for judging the approach to 0 is to integrate the buffer capacity values after the buffer capacity starts to decrease into a sequence, and to judge that the limit of the sequence is 0 according to the mathematical definition of the limit; For any jump batch, if the jump batch is a buffer capacity sudden drop batch, the jump batch is marked as a jump overlap batch; For any non-jump batch, if the non-jump batch is a buffer capacity sudden drop batch, the non-jump batch is marked as a non-jump overlap batch; For any jump overlap batch, the deviation of the pH jump time point and the buffer capacity sudden drop time point is calculated, and compared with a preset deviation, if the deviation is less than the preset deviation, and the pH jump time point is after the buffer capacity sudden drop time point, the jump overlap batch is marked as a buffer saturation batch; The proportion of the buffer saturation batch in the jump batch and the proportion of the non-jump overlap batch in the non-jump batch are respectively counted, to obtain a buffer saturation proportion and a non-jump overlap proportion; The buffer saturation proportion and the non-jump overlap proportion are respectively compared with a preset proportion, if the buffer saturation proportion is greater than the preset proportion, and the non-jump overlap proportion is less than the preset proportion, then the jump is caused by buffer saturation; If so, the acid / alkali addition amount of the buffer saturation batch and the pH are segmented and fitted to determine the buffer saturation point, and the stability of the buffer saturation point of each buffer saturation batch is analyzed to determine the buffer saturation point of the current fermentation system; The determination process of the buffer saturation point includes: The buffer capacity and the pH data of the buffer saturation batch are extracted, and for any buffer saturation batch: The pH value is taken as the independent variable x, and the corresponding buffer capacity is taken as the dependent variable y, a quadratic polynomial fitting is performed, the fitting equation is y=ax 2 +bx+c, the fitting parameters are solved by the least square method, and the fitting curve is obtained; A sliding window is set, and the fitting curve is traversed through the sliding window for point-by-point sliding; Wherein, the setting of the sliding window can be set according to the speed of the change of the buffer capacity with the pH, for example: For fast changing system (e.g. amino acid fermentation, the buffer capacity drops from 4.0 to 0.5 in 10 minutes when pH increases from 6.5 to 7.0), small window (2-3 points) is needed to ensure the start point of the drop can be captured (too large window will combine the drop section with the flat section, covering the mutation); For slow changing system (e.g. antibiotic fermentation, the buffer capacity drops from 3.0 to 0.5 in more than 1 hour), large window (5-6 points) is needed to avoid random fluctuation (e.g. single addition error) being misjudged as trend mutation; For each sliding window, the local slope within the sliding window is calculated by linear regression, and the slope change rate of adjacent sliding windows is calculated; The slope change rate of adjacent sliding windows is compared with the preset change rate, and if the slope change rate of adjacent sliding windows is greater than the preset change rate, the intersection point of adjacent sliding windows is taken as the initial segmentation point; Based on the initial segmentation point, the data is divided into two sections, and linear fitting is performed respectively to obtain two straight line equations; For any section of data, taking pH value as independent variable x and buffer capacity as dependent variable y, the linear regression equation is y=kx+b, where k is the slope and b is the intercept; The least square method is used to calculate k and b, and the fitting equation of the left section is C=k1x pH+b1, and the fitting equation of the right section is C=k2x pH+b2; In order to avoid the disconnection of the two straight lines, the buffer capacity C of the two straight lines at the segmentation point is forced to be equal, i.e. k1x pH0+b1=k2x pH0+b2; F test is performed on the two fitting equations to determine whether there is a statistically significant difference, which is as follows: The residual sum of squares of the two fitting equations and the residual sum of squares of the overall fitting are calculated respectively, and F statistic is calculated; According to the degrees of freedom df1=p part-p total, df2=n-p total, the critical value Fα of the F distribution table at the significance level α=0.05 is found; If F>Fα, there is a statistically significant difference between the two fitting equations; The ratio of the slopes of the two fitting equations, i.e. k2 / k1, is calculated and compared with the preset threshold; If the ratio of the slopes is greater than the preset threshold, and the F test determines that there is a statistically significant difference between the two slopes, the intersection point of the two straight lines is the buffer saturation point of the current fermentation system, which is obtained by solving the simultaneous equations; The simultaneous equations are k1x pH+b1=k2x pH+b2, and the pH value of the buffer saturation point is obtained by solving the equations, i.e. pH=(b2-b1) / (k1-k2). The obtained pH value is substituted into any one of the straight line equations to obtain the buffer capacity of the buffer saturation point; Calculate the mean value of all buffer saturation batch buffer saturation point buffer capacity as the buffer saturation point under the current fermentation system; Buffer saturation time prediction module: real-time monitoring and analysis of the buffer capacity trend of the current fermentation process, and combining the buffer saturation point of the current fermentation system, the buffer saturation time is predicted; The prediction process of the buffer saturation time includes: Real-time monitoring of the current fermentation process, the monitoring indicators include real-time pH value, acid / base addition amount, buffer concentration, cell concentration, metabolite concentration; Real-time buffer capacity is calculated after each monitoring, and a real-time buffer capacity time series is generated; Set a buffer capacity threshold, and compare the real-time buffer capacity with the buffer capacity threshold after each monitoring; If the current buffer capacity exceeds the buffer capacity threshold, calculate the difference between the current buffer capacity and the buffer saturation point to obtain the approach value; Compare the approach value with the preset threshold, if the approach value is less than or equal to the preset threshold, mark the time when the buffer capacity reaches the buffer capacity threshold as the critical time; Mark the time period between the critical time and the current time as the buffer warning window period, and integrate the buffer capacity monitoring values in the buffer warning window period into a buffer warning capacity monitoring sequence; Linear fitting is performed on the buffer warning capacity monitoring sequence, and the least squares method is used to solve the fitting equation, and the fitting goodness of the fitting equation is calculated, if the fitting goodness is greater than or equal to the empirical threshold, then the buffer warning capacity monitoring sequence is linear, otherwise, it is nonlinear; If the buffer warning capacity monitoring sequence is linear, linear fitting is performed on the buffer warning capacity monitoring sequence, and the least squares method is used to solve the fitting equation β(t)=kt+b; β(t)=the time when the buffer saturation point is the buffer saturation prediction time; If the buffer warning capacity monitoring sequence is nonlinear, the real-time monitoring index time series data of the fermentation process is taken as the core, including: buffer warning capacity monitoring sequence, real-time pH value, acid / base cumulative addition amount, buffer concentration, cell concentration, metabolite concentration; Align all monitoring indicators according to the timestamp, and standardize them; Extract historical information through sliding window, construct input-output sample pairs, including: Set the sliding window length L, that is, the length of the historical time series of the model input, where the sliding window length is set according to the average length of the buffer warning window period, for example: the average duration of the out-of-bound phase is 30min, take L=6 steps, that is, include 30min of historical data; Each input sample includes multivariate time series data within a sliding window, with dimensions (L, M), where L is the length of the sliding window and M is the total number of monitoring indicators; The output sample is the remaining time from the current time to the buffer saturation time; The structure of bidirectional LSTM+Dropout+fully connected layer is adopted to capture nonlinear time series features in layers and build an LSTM model; The input dimension is specified in the first LSTM layer, and the first layer LSTM outputs a complete time series sequence; The second layer LSTM outputs the final integrated feature vector; A Dropout layer is added after the LSTM layer to randomly discard 20% of the neuron connections, avoiding the model from over-relying on local features; The time series features extracted by LSTM are mapped to the predicted value of the remaining saturation time through a fully connected layer; Real-time monitoring data in the buffer warning window period is generated into input samples according to the sliding window, and features are automatically calculated and standardized. The input samples are input into the LSTM model, and the remaining saturation time is output; The current time and the remaining saturation time are summed to obtain the predicted buffer saturation time; The acid / base control amount calculation module: compares the predicted buffer saturation time with the fermentation end time. If the predicted saturation time is before the fermentation end time, the acid / base addition amount and the buffer capacity between the critical time of the buffer saturation batch and the fermentation end time are analyzed for correlation to obtain a correlation model. The target buffer capacity consumption rate is determined according to the buffer capacity at the current time, and the acid / base control amount is obtained by substituting the correlation model; The process of analyzing the correlation between the acid / base addition amount and the buffer capacity of the buffer saturation batch includes: Compare the predicted saturation time with the fermentation end time; If the predicted saturation time is after the fermentation end time, the buffer system will not be saturated before the fermentation ends, and no control is needed; On the contrary, the buffer system will be saturated before the fermentation ends, and control is needed to avoid pH out of control affecting the product. Specifically: Extract all key data of the buffer saturation batch, including: Time series data: acid / base cumulative addition amount in the fermentation process, buffer capacity at the corresponding time, fermentation time; Key node data: buffer saturation time of each batch, fermentation end time, acid / base addition amount at the buffer saturation time, acid / base addition amount at the fermentation end time; The acid / base cumulative addition amount of the fermentation process and the buffer capacity at the corresponding moment are normalized according to the relative time, the buffer saturation time of the buffer saturation batch is taken as the benchmark, the time is converted into the relative length of time from the saturation time, and the influence of the absolute time length difference of different batches is eliminated; The acid / base addition amount and the corresponding buffer capacity before and after each addition of acid / base in the buffer warning window period of the buffer saturation batch are extracted; For the buffer capacity of any buffer saturation batch, the time differentiation is carried out to obtain the instantaneous consumption rate: , wherein, is the monitoring interval, and the negative sign indicates that the buffer capacity decreases with time; Taking the instantaneous acid / base addition rate as the independent variable and the buffer capacity consumption rate as the dependent variable, a regression equation is fitted by using the least square method: If it is a linear relationship: r B =k·r A +b, k is the acid / base addition-buffer consumption correlation coefficient, and b is the basic consumption item caused by the acid / base produced by the metabolism of the cell; If it is a nonlinear relationship, wherein a is a quadratic term coefficient, reflecting the accelerated consumption effect of the buffer capacity under high addition rate; The residual buffer capacity is obtained by subtracting the buffer capacity at the buffer saturation point from the buffer capacity at the current time; The time period between the current time and the fermentation end time is marked as the allowed consumption time; The target buffer capacity consumption rate is obtained by proportional calculation of the residual buffer capacity and the allowed consumption time; It should be noted that the target buffer capacity consumption rate should be less than or equal to the average buffer capacity consumption rate between the critical time and the buffer saturation time in the historical batch, so as to avoid excessive limitation of the addition to cause pH out of control; The allowed acid / base addition rate and cumulative addition amount are obtained by substituting the target buffer capacity consumption rate into the correlation model.

[0022] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for pH control in a microbial fermentation process, characterized in that: Comprise: Step 1: The pH value of the acid / alkali before and after the addition of the same fermentation system under the history of different batches of microbial fermentation process is analyzed, and the history batch is divided into jump batch and non-jump batch, and the buffer capacity is verified to identify the buffer capacity jump batch, and the jump batch, non-jump batch and buffer capacity jump batch are analyzed respectively, and whether the pH value jump before and after the addition of acid / alkali is due to buffer saturation is judged; Step 2: If so, the acid / alkali addition amount and pH of the buffer saturation batch are segmented and fitted to determine the buffer saturation point, and the stability of the buffer saturation point of each buffer saturation batch is analyzed to determine the buffer saturation point of the current fermentation system; Step 3: The buffer capacity trend of the current fermentation process is monitored and analyzed in real time, and the buffer saturation time is predicted in combination with the buffer saturation point of the current fermentation system; Step 4: The predicted buffer saturation time is compared with the fermentation end time, if the predicted saturation time is before the fermentation end time, the acid / alkali addition amount and buffer capacity between the buffer saturation batch critical time and the fermentation end time are analyzed for correlation, the correlation model is obtained, and the target buffer capacity consumption rate is determined according to the current buffer capacity, and the acid / alkali control amount is obtained by substituting the correlation model.

2. The pH value regulation method of the microbial fermentation process according to claim 1, characterized in that: The jump batch and non-jump batch are divided as follows: The time point of each addition of acid / alkali in the microbial fermentation process and the corresponding pH value are obtained, and the pH change amount before and after the addition of acid / alkali is calculated, and the pH change amount time sequence is integrated in time sequence; Based on the normal batch historical data of the same fermentation system, the mean and standard deviation of the pH change amount in the normal fermentation stage are calculated, and the jump threshold is set based on the 3σ principle, wherein the normal batch is the historical batch with stable pH value in the fermentation process, which is judged by calculating the standard deviation of the pH value; If the pH change amount is greater than the jump threshold, the acid / alkali addition is marked as pH jump addition; In the pH change amount sequence, from the pH jump addition, the pH change amount is compared with the invalid threshold, if the pH change amount is less than the invalid threshold, the acid / alkali addition is marked as invalid addition; If there is invalid addition and pH jump addition, the historical batch is marked as jump batch, and the pH jump time point is recorded; On the contrary, if there is other situation except invalid addition and pH jump addition, it is non-jump batch.

3. The pH value regulation method of the microbial fermentation process according to claim 2, characterized in that: The identification method of the buffer capacity jump batch is as follows: According to the formula of buffer capacity, the acid / alkali concentration change amount before and after each addition is calculated in proportion with the pH change amount to obtain the buffer capacity, and the buffer capacity time sequence is integrated according to the addition time sequence; Based on the normal batch data of the same fermentation system, the normal buffer capacity interval is determined, and the buffer capacity in the buffer capacity time sequence is compared with the normal buffer capacity interval; If the buffer capacity time series all meet the sudden drop determination conditions, the historical batch is marked as a buffer capacity sudden drop batch, and the time point at which the buffer capacity first falls below the normal buffer capacity interval, i.e., the time point at which the buffer capacity begins to decrease, is recorded. The sudden drop determination conditions include: Determination condition one: any and more of the buffer capacity in the buffer capacity time series is below the normal buffer capacity interval; Determination condition two: the buffer capacity time series shows a continuous decrease; Determination condition three: the buffer capacity values after the buffer capacity begins to decrease are integrated into a sequence, and according to the mathematical definition of limit, it is judged that the limit of the sequence is 0.

4. The method for adjusting the pH value in a microbial fermentation process according to claim 2, characterized in that: The judgment method for whether the pH value jump before and after the acid / alkali addition is caused by buffer saturation is: For any jump batch, if the jump batch is a buffer capacity sudden drop batch, the jump batch is marked as a jump overlap batch; For any non-jump batch, if the non-jump batch is a buffer capacity sudden drop batch, the non-jump batch is marked as a non-jump overlap batch; For any jump overlap batch, the deviation of the pH jump time point and the buffer capacity sudden drop time point is calculated, and compared with a preset deviation, if the deviation is less than the preset deviation, and the pH jump time point is after the buffer capacity sudden drop time point, the jump overlap batch is marked as a buffer saturation batch; The proportion of the buffer saturation batches in the jump batches is counted to obtain a buffer saturation proportion; The proportion of the non-jump overlap batches in the non-jump batches is counted to obtain a non-jump overlap proportion; The buffer saturation proportion and the non-jump overlap proportion are compared with a preset proportion respectively, if the buffer saturation proportion is greater than the preset proportion, and the non-jump overlap proportion is less than the preset proportion, the jump is caused by buffer saturation.

5. The method for adjusting the pH value in a microbial fermentation process according to claim 1, characterized in that: The determination process of the buffer saturation point of the current fermentation system is: The buffer capacity and the pH data of the buffer saturation batches are extracted, and for any buffer saturation batch: The fitting curve is obtained by performing quadratic polynomial fitting on the pH value as the independent variable and the corresponding buffer capacity as the dependent variable, and solving the fitting parameters by the least square method; A sliding window is set, and the fitting curve is traversed by the sliding window for point-by-point sliding; For each sliding window, the local slope in the sliding window is calculated by linear regression, and the slope change rate of adjacent sliding windows is calculated; The slope change rate of adjacent sliding windows is compared with a preset change rate, if the slope change rate of adjacent sliding windows is greater than the preset change rate, the intersection point of adjacent sliding windows is taken as an initial segmentation point; Based on the initial segmentation point, the data is divided into two segments, and linear fitting is performed respectively to obtain two straight line equations; A sliding window is set, the linear regression slope in each window is calculated, and the slope change rate of adjacent windows is calculated; The slope change rate of adjacent windows is compared with a preset change rate, if the slope change rate of adjacent windows is greater than the preset change rate, the intersection point of adjacent windows is taken as an initial segmentation point; Based on the initial segmentation point, the data is divided into two segments, and linear fitting is performed respectively to obtain two straight line equations; F test is performed on the two fitting equations; The ratio of the slopes of the two fitting equations is calculated and compared with the preset threshold value; If the ratio of the slopes is greater than the preset threshold value, and the F test determines that there is a statistical difference between the slopes of the two segments, then the intersection point of the two straight lines is the buffer saturation point under the current fermentation system.

6. The method for adjusting the pH value of a microbial fermentation process according to claim 1, characterized in that: The prediction process of the buffer saturation time is as follows: Real-time monitoring of the current fermentation process is performed, and the buffer capacity is calculated in real time after each monitoring; A buffer capacity threshold value is set, and after each monitoring, the real-time buffer capacity is compared with the buffer capacity threshold value; If the current buffer capacity exceeds the buffer capacity threshold value, the difference between the current buffer capacity and the buffer saturation point is calculated to obtain an approximation value; If the approximation value is less than or equal to the preset threshold value, the time when the buffer capacity reaches the buffer capacity threshold value is marked as the critical time; The time period between the critical time and the current time is marked as the buffer warning window period, and the buffer capacity monitoring values within the buffer warning window period are integrated into a buffer warning capacity monitoring sequence; Linear fitting is performed on the buffer warning capacity monitoring sequence, the least squares method is used to solve the fitting equation, the goodness of fit of the fitting equation is calculated, and if the goodness of fit is greater than or equal to the empirical threshold value, the buffer warning capacity monitoring sequence is linear, otherwise it is nonlinear; If the buffer warning capacity monitoring sequence is linear, the time when the fitting equation is equal to the buffer saturation point is the predicted saturation time; If the buffer warning capacity monitoring sequence is nonlinear, a nonlinear prediction model is constructed based on the real-time monitoring index time series data of the fermentation process to obtain the predicted saturation time.

7. The method for adjusting the pH value of a microbial fermentation process according to claim 6, characterized in that: The way to construct a nonlinear prediction model to predict the buffer saturation time is as follows: The real-time monitoring index time series data of the fermentation process includes: buffer warning capacity monitoring sequence, real-time pH value, acid / base cumulative addition amount, buffer concentration, cell concentration, and metabolite concentration; All monitoring indexes are aligned according to the timestamp and standardized; Historical information is extracted through a sliding window to construct input-output sample pairs; A structure of bidirectional LSTM + Dropout + fully connected layer is adopted to capture nonlinear time series features in layers and construct an LSTM model; Real-time monitoring data in the buffer warning window period are input into the LSTM model to output the remaining saturation time. The buffer saturation prediction time is obtained by summing the current time and the remaining saturation time.

8. The method for adjusting the pH value of a microbial fermentation process according to claim 1, characterized in that: The process of correlation analysis of the acid / base addition amount and buffer capacity at the buffer saturation batch critical time and the fermentation end time is as follows: The acid / base cumulative addition amount of the fermentation process and the buffer capacity at the corresponding time are normalized according to the relative time, and the time is converted to the relative length of time from the saturation time, taking the buffer saturation time of the buffer saturation batch as the reference. The amount of acid / alkali added each time the buffer is supplemented within the buffer warning window and the buffer capacity before and after the addition of acid / alkali for each buffer saturation batch; For any buffer-saturated batch, the buffer capacity is time-differentiated to obtain the instantaneous consumption rate: where, is the monitoring interval, and the negative sign indicates that the buffer capacity decreases over time; Taking the instantaneous acid / alkali addition rate as the independent variable and the buffer capacity consumption rate as the dependent variable, a regression equation is fitted by the least squares method.

9. The method for regulating pH value in a microbial fermentation process according to claim 8, characterized in that: The determination of the target buffer capacity consumption rate and the acid / alkali control amount is as follows: The residual buffer capacity is obtained by subtracting the buffer capacity at the current time from the buffer capacity at the buffer saturation point; The time period between the current time and the fermentation end time is marked as the allowed consumption time; The target buffer capacity consumption rate is obtained by proportional calculation of the residual buffer capacity and the allowed consumption time; The allowed acid / alkali addition rate and cumulative addition amount are obtained by substituting the target buffer capacity consumption rate into the correlation model.

10. A pH regulation system for a microbial fermentation process, comprising The method comprises the following modules: A buffer saturation judgment module: the pH values before and after the addition of acid / alkali in the microbial fermentation process of different batches under the same fermentation system are analyzed for jump, the historical batches are divided into jump batches and non-jump batches, and the buffer capacity is verified to identify the buffer capacity jump batches, the jump batches, the non-jump batches and the buffer capacity jump batches are analyzed for overlap, and it is judged whether the pH value jump before and after the addition of acid / alkali is caused by buffer saturation; A buffer saturation point determination module: if yes, the acid / alkali addition amount and the pH of the buffer saturation batch are segmented and fitted to determine the buffer saturation point, and the stability of the buffer saturation points of the buffer saturation batches is analyzed to determine the buffer saturation point of the current fermentation system; A buffer saturation time prediction module: the trend of the buffer capacity of the current fermentation process is monitored and analyzed in real time, and the buffer saturation time is predicted in combination with the buffer saturation point of the current fermentation system; An acid / alkali control amount calculation module: the predicted buffer saturation time is compared with the fermentation end time, if the predicted saturation time is before the fermentation end time, the acid / alkali addition amount and the buffer capacity between the critical time of the buffer saturation batch and the fermentation end time are analyzed for correlation to obtain a correlation model, and the target buffer capacity consumption rate is determined according to the buffer capacity at the current time, and the acid / alkali control amount is obtained by substituting the target buffer capacity consumption rate into the correlation model.

Images