Intelligent fermentation tail gas treatment system

Abstract

The invention discloses an intelligent fermentation tail gas treatment system, and relates to the technical field of biological fermentation engineering, and the system comprises a data acquisition module which obtains fermentation tank biological metabolism parameters, gas inlet mass flow, a tail gas component spectrum and a tail gas treatment unit operation state in real time, and establishes a fermentation process-tail gas treatment complete period space-time synchronization industrial database; the health state prediction module learns different fermentation process batch pollution production characteristic rules, generates a process pollution production characteristic digital portrait, establishes a tail gas treatment system digital twinborn model, and predicts tail gas load fluctuation, a pollutant removal efficiency attenuation track and a treatment unit health state; and the stable up-to-standard emission module dynamically generates an optimal control instruction set based on the predicted tail gas load fluctuation, the pollutant removal efficiency attenuation track and the health state of the processing unit in combination with the process-pollutant production characteristic digital portrait, so that stable up-to-standard emission of the tail gas is ensured with the lowest energy consumption. The operation energy efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of bio-fermentation engineering technology, specifically to an intelligent system for treating fermentation exhaust gas. Background Technology

[0002] In existing technologies, fermentation exhaust gas treatment systems typically adopt an isolated control mode of polluting first and then treating. There is a lack of data linkage and collaborative optimization between the fermentation process and the exhaust gas treatment unit. They rely on manual experience or fixed thresholds for regulation, which makes it difficult to cope with complex operating conditions such as strong nonlinearity of the fermentation process, dynamic fluctuations in pollution load, and performance degradation of treatment equipment. This results in technical defects such as high energy consumption, poor emission stability, slow response, and inability to predict and prevent the risk of exceeding standards. Summary of the Invention

[0003] To solve the above-mentioned technical problems, an intelligent treatment system for fermentation exhaust gas is provided. This technical solution solves the aforementioned problems.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A fermentation exhaust gas intelligent treatment system, comprising: Data acquisition module, health status prediction module, and stable emission compliance module; Among them, the health status prediction module is electrically connected to the data acquisition module, and the stable emission compliance module is electrically connected to the health status prediction module. The data acquisition module, based on multi-source sensors, acquires in real time the biological metabolic parameters of the fermenter, the inlet gas mass flow rate, the exhaust gas composition spectrum and the operating status of the exhaust gas treatment unit, and establishes a spatiotemporally synchronized industrial database for the entire fermentation process and exhaust gas treatment cycle. The health status prediction module, based on the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle, learns the pollution generation characteristics of different fermentation process batches, generates a digital profile of process pollution generation characteristics, establishes a digital twin model of the exhaust gas treatment system, and predicts exhaust gas load fluctuations, pollutant removal efficiency decay trajectories, and the health status of the treatment unit. The stable emission compliance module dynamically generates the optimal control instruction set based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status, combined with the process-pollution generation characteristic digital profile, to ensure stable exhaust gas emission compliance with the lowest energy consumption.

[0005] Preferably, the data acquisition module specifically includes: The biological metabolic parameter unit, based on sensors deployed in the fermenter, collects biological metabolic parameters such as fermenter temperature, pH, dissolved oxygen (DO), tank pressure, turbidity, and carbon dioxide release rate in real time at a frequency of once per second. The intake air mass flow acquisition unit, based on a mass flow meter, acquires the intake air mass flow in real time; The exhaust gas composition spectrum acquisition unit, based on a Fourier transform infrared spectrometer, acquires the composition spectrum and concentration of VOCs, ethanol, ammonia and hydrogen sulfide pollutants in the exhaust gas every 10 seconds to obtain the exhaust gas composition spectrum.

[0006] Preferably, the data acquisition module also includes: The biofilter operation status unit acquires the real-time operation status of the biofilter based on temperature, humidity, bed pressure drop and inlet / outlet concentration sensors; The regenerative thermal oxidizer operation status unit acquires the real-time operation status of the regenerative thermal oxidizer based on combustion chamber temperature sensors, valve switching status sensors, fuel flow sensors, and inlet and outlet temperature sensors. The scrubbing tower operation status unit acquires the scrubbing tower operation status in real time based on sensors for circulating liquid pH, conductivity, liquid level, flow rate, and inlet / outlet concentration. The exhaust gas treatment unit operation status unit integrates the operation status of the biofilter, the regenerative thermal incinerator, and the scrubbing tower to obtain the exhaust gas treatment unit operation status.

[0007] Preferably, the data acquisition module also includes: The industrial database establishment unit transmits the fermenter's biological metabolic parameters, inlet gas mass flow rate, exhaust gas composition spectrum, and exhaust gas treatment unit's operating status to the industrial IoT gateway in real time via the MQTT protocol, unifies the time stamp alignment, and performs wavelet transform noise reduction on the data. The database establishment unit sets a fixed 30-minute time window, calculates the mean, standard deviation, maximum and minimum values ​​and trend slope of all data within the time window, obtains the statistical characteristics of all data within the time window, associates all data with the corresponding equipment metadata, stores them in the time series database, and establishes a spatiotemporally synchronized industrial database for the entire cycle of fermentation process and exhaust gas treatment. The device metadata includes: device ID, location, model, range, and calibration date.

[0008] Preferably, the health status prediction module specifically includes: The batch data unit obtains the process parameters, metabolic parameters and environmental parameters of different fermentation processes based on batch data of different fermentation processes in the historical database; The process parameters include: strain type, culture medium formula, batch number, aeration rate, stirring speed, tank pressure, and temperature; Metabolic parameters include: carbon dioxide release rate, dissolved oxygen concentration, pH value, and feed rate; Environmental parameters include: exhaust gas VOCs concentration, ethanol, ethyl acetate, instantaneous exhaust gas flow rate, and cumulative emissions; The metabolic intensity unit, based on the carbon dioxide release rate in the biological metabolic parameters, plots the carbon dioxide release rate curve, calculates the mean and slope of the carbon dioxide release rate within the window, and obtains the metabolic intensity of different fermentation batches. The instantaneous pollutant yield unit calculates the instantaneous pollutant yield of different fermentation processes based on the VOCs concentration in the exhaust gas and the instantaneous flow rate of the waste gas. The pollutant cumulative load unit calculates the total emission from the start of the time window to the current time based on the instantaneous yield of pollutants in the fermentation process batch and by time integration, thus obtaining the pollutant cumulative load of different fermentation process batches. The load fluctuation coefficient unit calculates the ratio of the standard deviation to the mean of the instantaneous yield of pollutants in a batch of fermentation process within 1 hour, thus obtaining the load fluctuation coefficient for different batches of fermentation process. The process stage feature unit automatically divides the fermentation process into stages based on the carbon dioxide release rate curve, namely the lag phase, logarithmic growth phase, stationary phase, and decline phase. It calculates the average value, peak value, and duration of each fermentation process stage to obtain the process stage characteristics of different fermentation batches. The feature matrix unit integrates the metabolic intensity, instantaneous pollutant yield, cumulative pollutant load, load fluctuation coefficient and process stage characteristics of different fermentation process batches to learn the pollution generation characteristics of different fermentation process batches and establish the pollution generation time series feature matrix of each fermentation process batch. The batch pollution generation feature unit of the fermentation process is based on the pollution generation time series feature matrix of each fermentation process batch, with unified timestamps, data preprocessing, principal component analysis algorithm is used to calculate the pollution generation feature covariance matrix of each fermentation process batch, and eigenvalue decomposition is performed to obtain the pollution generation feature values ​​and corresponding feature vectors of each fermentation process batch. The feature dataset unit selects the eigenvectors corresponding to the first m eigenvalues ​​as principal components based on the pollution generation feature values ​​of each fermentation process batch. The pollution generation time series feature matrix of each fermentation process batch is projected onto the principal components to obtain the dimensionality-reduced pollution generation time series feature dataset of each fermentation process batch.

[0009] Preferably, the health status prediction module also includes: The process pollution characteristic digital profile establishment unit, based on the pollution time series characteristic dataset of each fermentation process batch, uses the K-means clustering algorithm to randomly select K pollution characteristic data points of fermentation process batches as initial cluster centers. Each pollution characteristic data point of fermentation process batch is assigned to the nearest initial cluster center. The mean of all pollution characteristic data points of fermentation process batches in each cluster is recalculated. The cluster centers are iteratively updated until the cluster centers no longer change. All pollution characteristic data points of fermentation process batches are labeled, and the pollution characteristics of different fermentation process batches are classified into stable type, mid-term outbreak type and continuous high load type, generating a process pollution characteristic digital profile.

[0010] Preferably, the health status prediction module also includes: The degradation reference rate unit, based on literature on biofilters, obtains the reference values ​​of the maximum specific degradation rate and the half-saturation constant of biofilters, and calculates the reference rate of pollutant degradation in biofilters. The pollutant removal efficiency decay trajectory unit is based on real-time acquisition of the biofilter's operating status, including inlet concentration, empty bed residence time, and actual removal efficiency. It calculates the actual rate of pollutant degradation in the biofilter, establishes a biofilter dynamic model with the minimum value between the biofilter's pollutant degradation reference and actual rate as the objective, and predicts the pollutant removal efficiency decay trajectory of the biofilter. The regenerative thermal ignition (RTI) model unit calculates the heat generated when the exhaust gas is heated from the inlet temperature to the set incineration temperature, removes the heat recovered by the heat recovery unit, calculates the theoretical destruction rate under different VOCs components, concentrations and residence times, predicts the minimum incineration temperature and residence time of the RTI, and establishes the RTI model. The scrubbing tower absorption model unit is based on the real-time acquisition of the circulating liquid pH, conductivity, liquid level, flow rate and inlet and outlet concentrations of the scrubbing tower. Combined with the scrubbing tower design parameters and operating conditions, a scrubbing tower absorption model is established to predict the scrubbing tower's absorption efficiency for various pollutants and the change in absorbent saturation. The digital twin model unit of the exhaust gas treatment system integrates the biofilter kinetic model, the regenerative thermal incineration model, and the scrubbing tower absorption model to establish a digital twin model of the exhaust gas treatment system.

[0011] Preferably, the health status prediction module also includes: The performance degradation trend unit, based on the digital twin model of the exhaust gas treatment system, takes the runtime sequence data and pollution load data of each unit in the past N hours as input and the predicted operating trajectory of each unit in the next M hours as output to extract the performance degradation trend of the exhaust gas treatment system. The health status prediction unit, based on the known historical exhaust gas treatment system database, extracts the health status from the historical exhaust gas treatment system database according to manually recorded labels, and marks them as healthy, sub-healthy and requiring maintenance respectively. It trains an SVM support vector machine classifier and uses the performance degradation trend of the exhaust gas treatment system as input to predict the health status of the exhaust gas treatment unit in the next M hours. The tail gas load fluctuation trend unit takes the real-time biological metabolic parameters of the fermenter as input, combines them with the generated process pollution characteristic data profile, trains the machine learning model, and takes the real-time prediction of the tail gas generation and composition change trend during fermentation as output to extract the tail gas load fluctuation trend.

[0012] Preferably, the stable emission compliance module specifically includes: The optimal control instruction set unit, based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status as inputs, combines the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle with the digital twin model of the exhaust gas treatment system. It takes minimizing total energy consumption as the objective, sets the predicted emission exceedance risk as a penalty, and uses the allowable parameter range of the fermentation process and the safe operation boundary of the equipment in the industry standard as constraints. Using a nonlinear programming solver, it generates the optimal control instruction set for the entire fermentation process-exhaust gas treatment cycle in the future P hours through rolling time-domain optimization.

[0013] Preferably, the stable emission compliance module also includes: The unit for stable and compliant exhaust gas emission generates an optimal control instruction set for the entire lifecycle of fermentation process and exhaust gas treatment over a future P-hour period. This set is then sent to the execution layer, where real-time feedforward adjustment of the aeration ratio and stirring rate during fermentation stabilizes the pollution source. The unit also adaptively adjusts the nutrient solution addition to the biofilter, the combustion temperature and residence time of the regenerative thermal incinerator, and the circulation rate of the washing liquid in the exhaust gas treatment unit. This ensures stable and compliant exhaust gas emission with minimal energy consumption.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an intelligent treatment scheme for fermentation exhaust gas. This scheme constructs a real-time, synchronous, full-cycle industrial database through a data acquisition module, and utilizes a health status prediction module to learn process patterns and establish a digital twin model, achieving accurate prediction of exhaust gas load, treatment efficiency, and equipment health status. The stable emission compliance module can dynamically generate collaborative optimization instructions, breaking through the limitations of traditional isolated and static control. It achieves adaptive intelligent linkage between the fermentation pollution source and the exhaust gas treatment terminal, improving system operating efficiency and economy while strictly ensuring emission compliance and equipment safety, effectively preventing the risk of excessive emissions, and enhancing the overall stability and reliability of the treatment process. Attached Figure Description

[0015] Figure 1 This is a framework diagram of an intelligent fermentation exhaust gas treatment system. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] Reference Figure 1 As shown, a fermentation exhaust gas intelligent treatment system includes: Data acquisition module, health status prediction module, and stable emission compliance module; Among them, the health status prediction module is electrically connected to the data acquisition module, and the stable emission compliance module is electrically connected to the health status prediction module. The data acquisition module, based on multi-source sensors, acquires in real time the biological metabolic parameters of the fermenter, the inlet gas mass flow rate, the exhaust gas composition spectrum and the operating status of the exhaust gas treatment unit, and establishes a spatiotemporally synchronized industrial database for the entire fermentation process and exhaust gas treatment cycle. The health status prediction module, based on the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle, learns the pollution generation characteristics of different fermentation process batches, generates a digital profile of process pollution generation characteristics, establishes a digital twin model of the exhaust gas treatment system, and predicts exhaust gas load fluctuations, pollutant removal efficiency decay trajectories, and the health status of the treatment unit. The stable emission compliance module dynamically generates the optimal control instruction set based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status, combined with the process-pollution generation characteristic digital profile, to ensure stable exhaust gas emission compliance with the lowest energy consumption.

[0018] The data acquisition module specifically includes: The biological metabolic parameter unit, based on sensors deployed in the fermenter, collects biological metabolic parameters such as fermenter temperature, pH, dissolved oxygen (DO), tank pressure, turbidity, and carbon dioxide release rate in real time at a frequency of once per second. The intake air mass flow acquisition unit, based on a mass flow meter, acquires the intake air mass flow in real time; The exhaust gas composition spectrum acquisition unit, based on a Fourier transform infrared spectrometer, acquires the composition spectrum and concentration of VOCs, ethanol, ammonia and hydrogen sulfide pollutants in the exhaust gas every 10 seconds to obtain the exhaust gas composition spectrum.

[0019] The data acquisition module also includes: The biofilter operation status unit acquires the real-time operation status of the biofilter based on temperature, humidity, bed pressure drop and inlet / outlet concentration sensors; The regenerative thermal oxidizer operation status unit acquires the real-time operation status of the regenerative thermal oxidizer based on combustion chamber temperature sensors, valve switching status sensors, fuel flow sensors, and inlet and outlet temperature sensors. The scrubbing tower operation status unit acquires the scrubbing tower operation status in real time based on sensors for circulating liquid pH, conductivity, liquid level, flow rate, and inlet / outlet concentration. The exhaust gas treatment unit operation status unit integrates the operation status of the biofilter, the regenerative thermal incinerator, and the scrubbing tower to obtain the exhaust gas treatment unit operation status.

[0020] The data acquisition module also includes: The industrial database establishment unit transmits the fermenter's biological metabolic parameters, inlet gas mass flow rate, exhaust gas composition spectrum, and exhaust gas treatment unit's operating status to the industrial IoT gateway in real time via the MQTT protocol, unifies the time stamp alignment, and performs wavelet transform noise reduction on the data. The database establishment unit sets a fixed 30-minute time window, calculates the mean, standard deviation, maximum and minimum values ​​and trend slope of all data within the time window, obtains the statistical characteristics of all data within the time window, associates all data with the corresponding equipment metadata, stores them in the time series database, and establishes a spatiotemporally synchronized industrial database for the entire cycle of fermentation process and exhaust gas treatment. The device metadata includes: device ID, location, model, range, and calibration date.

[0021] When using it, combine the content of the above modules: Existing technologies for data acquisition in fermentation processes and exhaust gas treatment systems often suffer from problems such as scattered data sources, inconsistent acquisition frequencies, and inconsistent time scales. This makes it difficult to effectively integrate multi-source data, including fermenter biometabolism parameters, exhaust gas components, and treatment unit operating status. The raw data suffers from significant noise interference and lacks standardized processing, failing to form a high-quality industrial data foundation with full-cycle spatiotemporal synchronization, thus limiting the reliability of subsequent analysis and optimization applications. This step integrates multi-source sensor data, unifies time scale alignment, and performs wavelet transform denoising. Based on a fixed time window, statistical features are extracted, and the data is associated with and stored in a time-series database. This establishes an industrial database with full-cycle spatiotemporal synchronization for fermentation processes and exhaust gas treatment. Its beneficial effects include efficient fusion of multi-dimensional data and noise suppression, improving data accuracy and consistency, and providing a complete and reliable data foundation for real-time monitoring, trend analysis, and optimization decisions throughout the entire process.

[0022] The health status prediction module specifically includes: The batch data unit obtains the process parameters, metabolic parameters and environmental parameters of different fermentation processes based on batch data of different fermentation processes in the historical database; The process parameters include: strain type, culture medium formula, batch number, aeration rate, stirring speed, tank pressure, and temperature; Metabolic parameters include: carbon dioxide release rate, dissolved oxygen concentration, pH value, and feed rate; Environmental parameters include: exhaust gas VOCs concentration, ethanol, ethyl acetate, instantaneous exhaust gas flow rate, and cumulative emissions; The metabolic intensity unit, based on the carbon dioxide release rate in the biological metabolic parameters, plots the carbon dioxide release rate curve, calculates the mean and slope of the carbon dioxide release rate within the window, and obtains the metabolic intensity of different fermentation batches. The instantaneous pollutant yield unit calculates the instantaneous pollutant yield of different fermentation processes based on the VOCs concentration in the exhaust gas and the instantaneous flow rate of the waste gas. The pollutant cumulative load unit calculates the total emission from the start of the time window to the current time based on the instantaneous yield of pollutants in the fermentation process batch and by time integration, thus obtaining the pollutant cumulative load of different fermentation process batches. The load fluctuation coefficient unit calculates the ratio of the standard deviation to the mean of the instantaneous yield of pollutants in a batch of fermentation process within 1 hour, thus obtaining the load fluctuation coefficient for different batches of fermentation process. The process stage feature unit automatically divides the fermentation process into stages based on the carbon dioxide release rate curve, namely the lag phase, logarithmic growth phase, stationary phase, and decline phase. It calculates the average value, peak value, and duration of each fermentation process stage to obtain the process stage characteristics of different fermentation batches. The feature matrix unit integrates the metabolic intensity, instantaneous pollutant yield, cumulative pollutant load, load fluctuation coefficient and process stage characteristics of different fermentation process batches to learn the pollution generation characteristics of different fermentation process batches and establish the pollution generation time series feature matrix of each fermentation process batch. The batch pollution generation feature unit of the fermentation process is based on the pollution generation time series feature matrix of each fermentation process batch, with unified timestamps, data preprocessing, principal component analysis algorithm is used to calculate the pollution generation feature covariance matrix of each fermentation process batch, and eigenvalue decomposition is performed to obtain the pollution generation feature values ​​and corresponding feature vectors of each fermentation process batch. The feature dataset unit selects the eigenvectors corresponding to the first m eigenvalues ​​as principal components based on the pollution generation feature values ​​of each fermentation process batch. The pollution generation time series feature matrix of each fermentation process batch is projected onto the principal components to obtain the dimensionality-reduced pollution generation time series feature dataset of each fermentation process batch.

[0023] The health status prediction module also includes: The process pollution characteristic digital profile establishment unit, based on the pollution time series characteristic dataset of each fermentation process batch, uses the K-means clustering algorithm to randomly select K pollution characteristic data points of fermentation process batches as initial cluster centers. Each pollution characteristic data point of fermentation process batch is assigned to the nearest initial cluster center. The mean of all pollution characteristic data points of fermentation process batches in each cluster is recalculated. The cluster centers are iteratively updated until the cluster centers no longer change. All pollution characteristic data points of fermentation process batches are labeled, and the pollution characteristics of different fermentation process batches are classified into stable type, mid-term outbreak type and continuous high load type, generating a process pollution characteristic digital profile.

[0024] The health status prediction module also includes: The degradation reference rate unit, based on literature on biofilters, obtains the reference values ​​of the maximum specific degradation rate and the half-saturation constant of biofilters, and calculates the reference rate of pollutant degradation in biofilters. The pollutant removal efficiency decay trajectory unit is based on real-time acquisition of the biofilter's operating status, including inlet concentration, empty bed residence time, and actual removal efficiency. It calculates the actual rate of pollutant degradation in the biofilter, establishes a biofilter dynamic model with the minimum value between the biofilter's pollutant degradation reference and actual rate as the objective, and predicts the pollutant removal efficiency decay trajectory of the biofilter. The regenerative thermal ignition (RTI) model unit calculates the heat generated when the exhaust gas is heated from the inlet temperature to the set incineration temperature, removes the heat recovered by the heat recovery unit, calculates the theoretical destruction rate under different VOCs components, concentrations and residence times, predicts the minimum incineration temperature and residence time of the RTI, and establishes the RTI model. The scrubbing tower absorption model unit is based on the real-time acquisition of the circulating liquid pH, conductivity, liquid level, flow rate and inlet and outlet concentrations of the scrubbing tower. Combined with the scrubbing tower design parameters and operating conditions, a scrubbing tower absorption model is established to predict the scrubbing tower's absorption efficiency for various pollutants and the change in absorbent saturation. The digital twin model unit of the exhaust gas treatment system integrates the biofilter kinetic model, the regenerative thermal incineration model, and the scrubbing tower absorption model to establish a digital twin model of the exhaust gas treatment system.

[0025] The health status prediction module also includes: The performance degradation trend unit, based on the digital twin model of the exhaust gas treatment system, takes the runtime sequence data and pollution load data of each unit in the past N hours as input and the predicted operating trajectory of each unit in the next M hours as output to extract the performance degradation trend of the exhaust gas treatment system. The health status prediction unit, based on the known historical exhaust gas treatment system database, extracts the health status from the historical exhaust gas treatment system database according to manually recorded labels, and marks them as healthy, sub-healthy and requiring maintenance respectively. It trains an SVM support vector machine classifier and uses the performance degradation trend of the exhaust gas treatment system as input to predict the health status of the exhaust gas treatment unit in the next M hours. The tail gas load fluctuation trend unit takes the real-time biological metabolic parameters of the fermenter as input, combines them with the generated process pollution characteristic data profile, trains the machine learning model, and takes the real-time prediction of the tail gas generation and composition change trend during fermentation as output to extract the tail gas load fluctuation trend.

[0026] When using it, combine the content of the above modules: Current fermentation tail gas management relies heavily on manual experience and single-point parameter monitoring, lacking systematic quantitative analysis of the dynamic characteristics of waste gas generation in fermentation batches. Furthermore, tail gas treatment system operation optimization is often independent of upstream processes, making it difficult for treatment units to match the time-varying and complex nature of waste gas loads. This hinders coordinated early warning and precise control of the process and treatment processes, resulting in risks such as low treatment efficiency, high energy consumption, and delayed maintenance. This step integrates multi-source data from fermentation batches to construct a digital profile of waste gas generation characteristics and a digital twin model of the tail gas treatment system. This enables dynamic analysis of waste gas generation patterns and coordinated prediction of treatment unit performance degradation, helping to identify process anomalies and treatment unit health status in advance. It provides data-driven decision support for process optimization and precise control of the tail gas treatment system, improving treatment efficiency and operational stability.

[0027] The stable emission compliance module specifically includes: The optimal control instruction set unit, based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status as inputs, combines the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle with the digital twin model of the exhaust gas treatment system. It takes minimizing total energy consumption as the objective, sets the predicted emission exceedance risk as a penalty, and uses the allowable parameter range of the fermentation process and the safe operation boundary of the equipment in the industry standard as constraints. Using a nonlinear programming solver, it generates the optimal control instruction set for the entire fermentation process-exhaust gas treatment cycle in the future P hours through rolling time-domain optimization.

[0028] The stable emission compliance module also includes: The unit for stable and compliant exhaust gas emission generates an optimal control instruction set for the entire lifecycle of fermentation process and exhaust gas treatment over a future P-hour period. This set is then sent to the execution layer, where real-time feedforward adjustment of the aeration ratio and stirring rate during fermentation stabilizes the pollution source. The unit also adaptively adjusts the nutrient solution addition to the biofilter, the combustion temperature and residence time of the regenerative thermal incinerator, and the circulation rate of the washing liquid in the exhaust gas treatment unit. This ensures stable and compliant exhaust gas emission with minimal energy consumption.

[0029] When using it, combine the content of the above modules: In existing technologies, fermentation exhaust gas treatment typically employs independent, static control strategies, failing to dynamically and collaboratively optimize the fermentation process and exhaust gas treatment system. This lack of predictive control over exhaust gas load fluctuations, equipment efficiency degradation, and the risk of exceeding emission standards results in high energy consumption or insufficient emission stability. This approach constructs a spatiotemporally synchronized industrial database and digital twin model covering the entire fermentation process and exhaust gas treatment cycle. By dynamically generating collaborative control commands through rolling time-domain optimization, it achieves stabilization of pollution sources and adaptive adjustment of exhaust gas treatment unit parameters. Under the premise of strictly meeting emission standards and equipment safety constraints, it reduces total system energy consumption, effectively prevents the risk of exceeding emission standards, and improves overall operational stability and economy.

[0030] Based on the above, the specific implementation method is as follows: Taking the odor control project in the saccharification and fermentation workshop of a large brewery as an example, the data acquisition module constructed a panoramic perception network covering the source and the end. At the source, the system installed a high-precision sensor array in each 500-ton conical fermentation tank. Temperature probes monitored the gradient cooling of the fermentation process from 12℃ to 5℃; pH meters tracked changes in wort acidity; and online infrared carbon dioxide analyzers on the top of the tank captured the concentration of carbon dioxide produced during fermentation once per second, accurately calculated the release rate, and used it as a core metabolic indicator reflecting the main fermentation, diacetyl reduction, and wine storage. At the end, a Fourier transform infrared gas analyzer installed on the mixed exhaust gas main performs a full-spectrum scan of the exhaust gas every 10 seconds, accurately detecting the concentration spectrum of characteristic VOCs, mainly ethanol, acetaldehyde, and organic acids. It also collects key operating parameters in real time from downstream bio-trickling filter, regenerative thermal incinerator, and water scrubbing tower. These parameters include the temperature and humidity of the packing layer and the conductivity of the circulating liquid in the bio-trickling filter; the temperature of the combustion chamber in the regenerative thermal incinerator (6 temperature zones), the status of the regenerative switching valve, and the natural gas flow rate; and the frequency and level of the circulating water pump in the water scrubbing tower. All sensor data from different brands and protocols are also collected. Protocol conversion is performed at the edge IoT gateway, and the data is uploaded to the cloud industrial IoT platform using the MQTT protocol. The platform uses the Network Time Protocol to stamp all data streams with a unified timestamp, and uses wavelet transform algorithm to denoise the raw data. The system uses a fixed window of 30 minutes to calculate the statistical characteristics of all raw data within the window, such as the mean, standard deviation, maximum value, and trend slope. After associating these characteristics with the device metadata of each sensor, the data is stored in the time series database to establish a spatiotemporally synchronized industrial database that integrates the entire process of beer fermentation process status, characteristic pollutant generation, and multi-stage exhaust gas treatment efficiency. Based on a spatiotemporally synchronized industrial database that integrates the entire process of beer fermentation—from the state of the process to the generation of characteristic pollutants and the efficiency of multi-stage exhaust gas treatment—the health status prediction module begins to perform in-depth analysis. The system first extracts features from fermentation batch data of hundreds of different beer varieties over the past year. For a typical pale lager, the system automatically plots a release rate curve, identifying the phase characteristics of a rapid increase in release rate during the main fermentation period, a stable release rate during the diacetyl reduction period, and a gradual decrease in release rate during the storage period. Combined with the instantaneous ethanol yield monitored by a Fourier transform infrared spectroscopy gas analyzer, the system quantifies the pollution intensity and volatility of each stage. After reducing the dimensionality of multiple feature dimensions such as the mean release rate, peak ethanol yield, load fluctuation coefficient, and duration of each stage through principal component analysis, the system successfully classified historical batches into three typical process pollution characteristics digital profiles using the K-means clustering algorithm: stable type, rapid outbreak type, and continuous medium load type. The module establishes a digital twin model of the exhaust gas treatment system based on a fusion of mechanism and data-driven approaches. The bio-trickling filter kinetic model simulates the degradation process of ethanol, acetaldehyde, organic acids, and nitrogen-containing compounds by heterotrophic bacteria and ammonia-oxidizing bacteria based on inlet gas composition, temperature, humidity, and empty bed residence time, and predicts the efficiency decline trajectory caused by biofilm thickening or nutrient imbalance. The regenerative thermal incinerator thermodynamic model calculates the heat recovery efficiency of the regenerator based on the real-time inlet gas calorific value, VOCs concentration spectrum, flow rate, and inlet temperature, and calculates the minimum amount of supplementary fuel required to ensure a destruction rate of over 99% and a residence time of more than 1 second. The water scrubbing tower mass transfer model predicts its absorption efficiency for water-soluble organic matter and acidic gases. Combining real-time fermentation data and matching pollution profiles, the module utilizes a trained long short-term memory neural network model to predict the fluctuation trend of total VOCs concentration in exhaust gas 3 hours in advance. The system uses a support vector machine classifier to pre-diagnose the health status of the treatment unit. When the unit pressure drop removal efficiency of the bio-trickling filter tower continues to deviate from the model prediction value and the bed temperature rises abnormally, the system will issue an early warning that the bio-filter bed is at risk of blockage or acidification and that the health status is sub-healthy. Faced with the predicted information, the stable emission compliance module executes closed-loop collaborative optimization control. The system predicts that the currently fermenting rapid burst batch will enter the main fermentation peak within the next 2 hours, and the VOCs load will surge by 50%. The efficiency of the regenerative thermal incinerator model has slightly decreased due to long-term operation, and the predicted health state is sub-healthy. The optimal control instruction set unit is immediately activated. With the next 4 hours as the optimization interval, a multi-objective nonlinear programming model is established. The objective function is to minimize the total energy consumption of the system, including the power consumption of fermenter stirring and aeration, RTO natural gas consumption, water washing and circulating water pump power consumption. The predicted emission concentration exceeding the standard is set as a high-weight penalty term. The constraints are the allowable range of fermentation process parameters, the maximum safe temperature of the regenerative thermal incinerator combustion chamber, and the upper limit of the processing capacity of each device. Using a sequential quadratic programming solver for rolling time-domain optimization calculations, the system generates and issues a set of collaborative instructions within seconds. At the source, it fine-tunes the aeration ratio and stirring speed of the fermenter, moderately smoothing the instantaneous release rate of volatile metabolites without affecting fermentation quality, thus achieving peak reduction at the source. At the processing side, it executes adaptive regulation, instructing the regenerative thermal incinerator to raise the baseline incineration temperature from 850℃ to 860℃ in advance and optimizing the valve switching cycle to compensate for the slight decrease in the efficiency of the regenerative medium, ensuring a high destruction rate. It also instructs the bio-trickling filter to increase the pulsed dosing frequency of the nutrient solution to maintain the activity of microorganisms under load shocks; and instructs the water washing tower to moderately increase the circulating liquid flow rate to enhance its buffering capacity against water-soluble shock loads. By employing a dynamic optimization strategy that combines feedforward and feedback, and coordinates source and end-point emissions, the system successfully kept the total concentration of exhaust emissions during peak periods below the limits set by the "Integrated Emission Standard for Air Pollutants". Compared with the traditional independent PID control strategy, this approach reduces overall energy consumption, effectively avoids treatment failures or energy waste caused by load shocks, and achieves a dual optimization of environmental compliance and operational economy.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A fermentation off-gas intelligent treatment system, characterized in that, include: Data acquisition module, health status prediction module, and stable emission compliance module; Among them, the health status prediction module is electrically connected to the data acquisition module, and the stable emission compliance module is electrically connected to the health status prediction module. The data acquisition module, based on multi-source sensors, acquires in real time the biological metabolic parameters of the fermenter, the inlet gas mass flow rate, the exhaust gas composition spectrum and the operating status of the exhaust gas treatment unit, and establishes a spatiotemporally synchronized industrial database for the entire fermentation process and exhaust gas treatment cycle. The health status prediction module, based on the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle, learns the pollution generation characteristics of different fermentation process batches, generates a digital profile of process pollution generation characteristics, establishes a digital twin model of the exhaust gas treatment system, and predicts exhaust gas load fluctuations, pollutant removal efficiency decay trajectories, and the health status of the treatment unit. The stable emission compliance module dynamically generates the optimal control instruction set based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status, combined with the process-pollution generation characteristic digital profile, to ensure stable exhaust gas emission compliance with the lowest energy consumption.

2. The intelligent system for processing fermentation off-gas according to claim 1, wherein, The data acquisition module specifically includes: The biological metabolic parameter unit, based on sensors deployed in the fermenter, collects biological metabolic parameters such as fermenter temperature, pH, dissolved oxygen (DO), tank pressure, turbidity, and carbon dioxide release rate in real time at a frequency of once per second. The intake air mass flow acquisition unit, based on a mass flow meter, acquires the intake air mass flow in real time; The exhaust gas composition spectrum acquisition unit, based on a Fourier transform infrared spectrometer, acquires the composition spectrum and concentration of VOCs, ethanol, ammonia and hydrogen sulfide pollutants in the exhaust gas every 10 seconds to obtain the exhaust gas composition spectrum.

3. The intelligent system for processing fermentation off-gas according to claim 2, wherein, The data acquisition module also includes: The biofilter operation status unit acquires the real-time operation status of the biofilter based on temperature, humidity, bed pressure drop and inlet / outlet concentration sensors; The regenerative thermal oxidizer operation status unit acquires the real-time operation status of the regenerative thermal oxidizer based on combustion chamber temperature sensors, valve switching status sensors, fuel flow sensors, and inlet and outlet temperature sensors. The scrubbing tower operation status unit acquires the scrubbing tower operation status in real time based on sensors for circulating liquid pH, conductivity, liquid level, flow rate, and inlet / outlet concentration. The exhaust gas treatment unit operation status unit integrates the operation status of the biofilter, the regenerative thermal incinerator, and the scrubbing tower to obtain the exhaust gas treatment unit operation status.

4. The intelligent system for processing fermentation off-gas according to claim 3, wherein, The data acquisition module also includes: The industrial database establishment unit transmits the fermenter's biological metabolic parameters, inlet gas mass flow rate, exhaust gas composition spectrum, and exhaust gas treatment unit's operating status to the industrial IoT gateway in real time via the MQTT protocol, unifies the time stamp alignment, and performs wavelet transform noise reduction on the data. The database establishment unit sets a fixed 30-minute time window, calculates the mean, standard deviation, maximum and minimum values ​​and trend slope of all data within the time window, obtains the statistical characteristics of all data within the time window, associates all data with the corresponding equipment metadata, stores them in the time series database, and establishes a spatiotemporally synchronized industrial database for the entire cycle of fermentation process and exhaust gas treatment. The device metadata includes: device ID, location, model, range, and calibration date.

5. The intelligent system for processing fermentation off-gas according to claim 1, wherein, The health status prediction module specifically includes: The batch data unit obtains the process parameters, metabolic parameters and environmental parameters of different fermentation processes based on batch data of different fermentation processes in the historical database; The process parameters include: strain type, culture medium formula, batch number, aeration rate, stirring speed, tank pressure, and temperature; Metabolic parameters include: carbon dioxide release rate, dissolved oxygen concentration, pH value, and feed rate; Environmental parameters include: exhaust gas VOCs concentration, ethanol, ethyl acetate, instantaneous exhaust gas flow rate, and cumulative emissions; The metabolic intensity unit, based on the carbon dioxide release rate in the biological metabolic parameters, plots the carbon dioxide release rate curve, calculates the mean and slope of the carbon dioxide release rate within the window, and obtains the metabolic intensity of different fermentation batches. The instantaneous pollutant yield unit calculates the instantaneous pollutant yield of different fermentation processes based on the VOCs concentration in the exhaust gas and the instantaneous flow rate of the waste gas. The pollutant cumulative load unit calculates the total emission from the start of the time window to the current time based on the instantaneous yield of pollutants in the fermentation process batch and by time integration, thus obtaining the pollutant cumulative load of different fermentation process batches. The load fluctuation coefficient unit calculates the ratio of the standard deviation to the mean of the instantaneous yield of pollutants in a batch of fermentation process within 1 hour, thus obtaining the load fluctuation coefficient for different batches of fermentation process. The process stage feature unit automatically divides the fermentation process into stages based on the carbon dioxide release rate curve, namely the lag phase, logarithmic growth phase, stationary phase, and decline phase. It calculates the average value, peak value, and duration of each fermentation process stage to obtain the process stage characteristics of different fermentation batches. The feature matrix unit integrates the metabolic intensity, instantaneous pollutant yield, cumulative pollutant load, load fluctuation coefficient and process stage characteristics of different fermentation process batches to learn the pollution generation characteristics of different fermentation process batches and establish the pollution generation time series feature matrix of each fermentation process batch. The batch pollution generation feature unit of the fermentation process is based on the pollution generation time series feature matrix of each fermentation process batch, with unified timestamps, data preprocessing, principal component analysis algorithm is used to calculate the feature covariance matrix of each fermentation process batch pollution generation, and eigenvalue decomposition is performed to obtain the pollution generation feature values ​​and corresponding feature vectors of each fermentation process batch. The feature dataset unit selects the eigenvectors corresponding to the first m eigenvalues ​​as principal components based on the pollution generation feature values ​​of each fermentation process batch. The pollution generation time series feature matrix of each fermentation process batch is projected onto the principal components to obtain the dimensionality-reduced pollution generation time series feature dataset of each fermentation process batch.

6. The intelligent fermentation tail gas treatment system according to claim 5, characterized in that, The health status prediction module also includes: The process pollution characteristic digital profile establishment unit, based on the pollution time series characteristic dataset of each fermentation process batch, uses the K-means clustering algorithm to randomly select K pollution characteristic data points of fermentation process batches as initial cluster centers. Each pollution characteristic data point of fermentation process batch is assigned to the nearest initial cluster center. The mean of all pollution characteristic data points of fermentation process batches in each cluster is recalculated. The cluster centers are iteratively updated until the cluster centers no longer change. All pollution characteristic data points of fermentation process batches are labeled, and the pollution characteristics of different fermentation process batches are classified into stable type, mid-term outbreak type and continuous high load type, generating a process pollution characteristic digital profile.

7. The intelligent fermentation tail gas treatment system according to claim 6, characterized in that, The health status prediction module also includes: The degradation reference rate unit, based on literature on biofilters, obtains the reference values ​​of the maximum specific degradation rate and the half-saturation constant of biofilters, and calculates the reference rate of pollutant degradation in biofilters. The pollutant removal efficiency decay trajectory unit is based on real-time acquisition of the biofilter's operating status, including inlet concentration, empty bed residence time, and actual removal efficiency. It calculates the actual rate of pollutant degradation in the biofilter, establishes a biofilter dynamic model with the minimum value between the biofilter's pollutant degradation reference and actual rate as the objective, and predicts the pollutant removal efficiency decay trajectory of the biofilter. The regenerative thermal ignition (RTI) model unit calculates the heat generated when the exhaust gas is heated from the inlet temperature to the set incineration temperature, removes the heat recovered by the heat recovery unit, calculates the theoretical destruction rate under different VOCs components, concentrations and residence times, predicts the minimum incineration temperature and residence time of the RTI, and establishes the RTI model. The scrubbing tower absorption model unit is based on the real-time acquisition of the circulating liquid pH, conductivity, liquid level, flow rate and inlet and outlet concentrations of the scrubbing tower. Combined with the scrubbing tower design parameters and operating conditions, a scrubbing tower absorption model is established to predict the scrubbing tower's absorption efficiency for various pollutants and the change in absorbent saturation. The digital twin model unit of the exhaust gas treatment system integrates the biofilter kinetic model, the regenerative thermal incineration model, and the scrubbing tower absorption model to establish a digital twin model of the exhaust gas treatment system.

8. The intelligent fermentation tail gas treatment system according to claim 7, characterized in that, The health status prediction module also includes: The performance degradation trend unit, based on the digital twin model of the exhaust gas treatment system, takes the runtime sequence data and pollution load data of each unit in the past N hours as input and the predicted operating trajectory of each unit in the next M hours as output to extract the performance degradation trend of the exhaust gas treatment system. The health status prediction unit, based on the known historical exhaust gas treatment system database, extracts the health status from the historical exhaust gas treatment system database according to manually recorded labels, and marks them as healthy, sub-healthy and requiring maintenance respectively. It trains an SVM support vector machine classifier and uses the performance degradation trend of the exhaust gas treatment system as input to predict the health status of the exhaust gas treatment unit in the next M hours. The tail gas load fluctuation trend unit takes the real-time biological metabolic parameters of the fermenter as input, combines them with the generated process pollution characteristic data profile, trains the machine learning model, and takes the real-time prediction of the tail gas generation and composition change trend during fermentation as output to extract the tail gas load fluctuation trend.

9. The intelligent fermentation tail gas treatment system according to claim 8, characterized in that, The stable emission compliance module specifically includes: The optimal control instruction set unit, based on the predicted exhaust gas load fluctuations, pollutant removal efficiency decay trajectory, and treatment unit health status as inputs, combines the spatiotemporally synchronized industrial database of the entire fermentation process-exhaust gas treatment cycle with the digital twin model of the exhaust gas treatment system. It takes minimizing total energy consumption as the objective, sets the predicted emission exceedance risk as a penalty, and uses the allowable parameter range of the fermentation process and the safe operation boundary of the equipment in the industry standard as constraints. Using a nonlinear programming solver, it generates the optimal control instruction set for the entire fermentation process-exhaust gas treatment cycle in the future P hours through rolling time-domain optimization.

10. The intelligent fermentation tail gas treatment system according to claim 9, characterized in that, The stable emission compliance module also includes: The unit for stable and compliant exhaust gas emission generates an optimal control instruction set for the entire lifecycle of fermentation process and exhaust gas treatment over a future P-hour period. This set is then sent to the execution layer, where real-time feedforward adjustment of the aeration ratio and stirring rate during fermentation stabilizes the pollution source. The unit also adaptively adjusts the nutrient solution addition to the biofilter, the combustion temperature and residence time of the regenerative thermal incinerator, and the circulation rate of the washing liquid in the exhaust gas treatment unit. This ensures stable and compliant exhaust gas emission with minimal energy consumption.

Images