Model predictive control based intelligent control system and method for egsb reactor

Abstract

The application discloses an EGSB reactor intelligent control system and method based on model prediction control. The application is based on the main structure of the EGSB reactor, and data acquisition modules, data storage modules and simulation control modules are constructed. The data acquisition modules are used for collecting water quality data of inlet and outlet water of the EGSB reactor and operation data of the inlet water pump and the reflux pump. The data storage modules are used for aggregating multi-source data from the data acquisition modules through a gateway device, simultaneously performing time series data preprocessing, and storing to a data warehouse. The simulation control modules are used for sending a query request to the data warehouse through a communication interface and accepting returned data, utilizing an EGSB reactor operation simulation prediction model and a control action optimization algorithm, generating output signals through calculation, and transmitting the output signals to the inlet water pump and the reflux pump through a control foot.

Description

Technical Field

[0001] This invention relates to the field of wastewater biological treatment technology, specifically to an intelligent control system and method for an expanded granular sludge bed (EGSB) reactor based on model predictive control. Background Technology

[0002] Anaerobic granular sludge technology has many advantages, such as high wastewater treatment efficiency, long sludge retention time, good sludge-water separation effect, and small reactor footprint, and is currently widely used in the treatment of high-concentration organic industrial wastewater. However, the growth of anaerobic microorganisms is easily affected by toxic substances, and the balance among anaerobic symbiotic bacteria is fragile. Anaerobic reactors still face engineering challenges such as toxic stress, easy acidification instability, and low methane yield.

[0003] Enhancing the stable operation of anaerobic reactors has long been a research hotspot in the field of biological wastewater treatment. In recent years, with the development of new-generation information technologies such as IoT sensing, big data, and artificial intelligence, strategies for improving reactor performance can be explored from the perspective of intelligent control.

[0004] The following existing technologies were found through a search: The patent specification with publication number CN110357236A discloses a smart control method for wastewater treatment plants based on a catastrophe inversion effluent prediction model. The method involves preprocessing data collected by equipment through a smart sensing module, inputting the predicted effluent water quality parameters into the catastrophe inversion model, and then inputting these predicted parameters into the smart control module. This generates corresponding equipment control commands, which are sent to the relevant equipment to achieve automatic operation. Simultaneously, a database records influent and effluent water quality data, predicted effluent water quality data, and corresponding control commands, automatically correcting model parameters to achieve a smart closed-loop system.

[0005] The patent specification with publication number CN119828482A discloses a short-range denitrification anaerobic ammonia oxidation system and its intelligent control method, including the following steps: constructing and processing a machine learning dataset to obtain a training set and a test set; based on the training set, establishing a prediction model through machine learning methods, evaluating the prediction performance to select the optimal prediction model; using the data in the dataset as feature data for importance analysis to identify key features; using the key features as optimization control parameters, constructing the optimal control strategy through particle swarm optimization algorithm. Summary of the Invention

[0006] The purpose of this invention is to maximize reactor performance. It provides an intelligent control system and method for an EGSB reactor based on model predictive control. Based on the sensor monitoring of reactor influent and effluent water quality and big data on reactor operation, a simulation and prediction model for EGSB reactor operation is constructed using deep learning algorithms. This model enables multi-step prediction of the reactor's future effluent water quality. Based on model predictive control theory, a rolling search is performed on the reactor's influent and reflux flow rates. This ensures that the concentrations of toxic characteristic pollutants, free volatile fatty acids (VFAs), and pH value in the reactor are all within their respective threshold ranges, thereby maximizing the removal load of organic matter and toxic characteristic pollutants in the reactor.

[0007] The specific technical solution of this invention is as follows: In a first aspect, the present invention provides an intelligent control system for an EGSB reactor based on model predictive control, which is based on the main structure of the EGSB reactor and includes a data acquisition module, a data storage module and a simulation control module. The data acquisition module is used to collect water quality data of the influent and effluent of the EGSB reactor, as well as operating data of the influent pump and return pump; The data storage module is used to aggregate multi-source data from the data acquisition module through the gateway device, perform time-series data preprocessing, and store the data in the data warehouse. The simulation control module is used to send query requests to the data warehouse through the communication interface and receive returned data. It uses the EGSB reactor to run the simulation prediction model and control action optimization algorithm, generates output signals through calculation, and transmits the output signals to the inlet pump and return pump through the control foot.

[0008] In some preferred embodiments, the data acquisition module collects the influent and effluent water quality data of the EGSB reactor from influent and effluent water quality monitoring sensors and water quality analyzers.

[0009] In some preferred embodiments, the influent and effluent water quality monitoring sensors include pH sensors and conductivity sensors, both of which are installed before the influent pump and at the EGSB reactor outlet.

[0010] In some preferred embodiments, the water quality analyzer includes a COD analyzer, an alkalinity analyzer, a volatile fatty acid analyzer, and a toxic characteristic pollutant analyzer. The volatile fatty acid analyzer's water intake is only located at the EGSB reactor outlet, while the COD analyzer, alkalinity analyzer, and toxic characteristic pollutant analyzer's water intakes are all located before the influent pump and at the EGSB reactor outlet.

[0011] In some preferred embodiments, the operating data for the inlet and return pumps include pump speed, pump inlet and outlet water pressure difference, and pump power consumption.

[0012] In some preferred embodiments, the pump flow rate is calculated based on the pump speed, the pressure difference between the pump inlet and outlet, and the pump power consumption, as specifically expressed below: In the formula, It is the pump's flow rate. It is the pump speed. It is the pressure difference between the pump inlet and outlet. It's the pump's power consumption. , , , It is the characteristic curve value of the pump.

[0013] In some preferred embodiments, the data acquired by the data acquisition module is transmitted via wired connection or wireless communication network.

[0014] In some preferred embodiments, time series data preprocessing extracts time series data based on a preset time resolution (e.g., 30 min) and applies... The criteria remove outliers and simultaneously fill in the gaps using an interpolation polynomial. Specifically, this can be represented as follows: In the formula, Represents probability. Represents the mean. Represents standard deviation, Represents each original observation in the dataset. Represents the moment Estimation of missing values, Represents the smoothing coefficient, 0 < <1, Representative moment The original observation value at the previous moment, Representative moment The estimated value at the previous moment.

[0015] In some preferred embodiments, the analog control module is a microcontroller system.

[0016] In some preferred embodiments, both the inlet pump and the return pump are variable frequency pumps.

[0017] In some preferred examples, The EGSB reactor operation simulation and prediction model constructs a sub-model for each effluent water quality indicator. The inputs are time-series data including current and historical influent and effluent water quality, influent pump flow rate, and return pump flow rate. The length of the historical time series is... It should be greater than the hydraulic retention time (HRT) of the EGSB reactor, where, at the current time The influent pump flow rate and return pump flow rate are determined by using a control action optimization algorithm to find the optimal flow rate values ​​and then applied to the operation of the EGSB reactor. ~ Within the time step; the model output is the next time step for the EGSB reactor. The water quality indicators at any given time; each sub-model is built on a convolutional neural network-gated recurrent unit-attention mechanism (CNN-GRU-Attention) architecture, and the hyperparameters are tuned using a Bayesian optimization algorithm to obtain the optimal model, which can be represented as follows: In the formula, for The inflow index data matrix is ​​constantly input into the model. , , , , , , Represent to The influent COD, toxic characteristic pollutants, alkalinity, and time index of the EGSB reactor at specific times (example, time series data with a time resolution of 30 min). Values ​​can range from 1 to 48), pH, conductivity, and flow rate data; for Input the effluent recirculation index data matrix into the model at any time. , , , , , , Represent to Real-time data on COD, toxic characteristic pollutants, alkalinity, VFAs, pH, conductivity, and reflux flow rate of the EGSB reactor effluent; For model output Real-time water quality index data matrix , , , , , Represent Model-predicted values ​​of COD, toxic characteristic pollutants, alkalinity, VFAs, pH, and conductivity of the effluent from the EGSB reactor at any given time. It is an EGSB reactor operation simulation and prediction model. , , , , , These are the prediction sub-models for COD, toxic characteristic pollutants, alkalinity, VFAs, pH, and conductivity in the EGSB reactor.

[0018] In some preferred embodiments, the control action optimization algorithm is based on the fundamental theory of model predictive control (MPC) and sets an objective function, specifically for the EGSB reactor. to The weighted sum of the removal loads of COD and toxic characteristic pollutants within a given time period must satisfy the principle of maximizing the objective function when performing optimization, which can be specifically expressed as follows: In the formula, and They are respectively to The removal load of COD and toxic characteristic pollutants in the EGSB reactor at a given time. , They are respectively Real-time monitoring data of COD and toxic characteristic pollutants in the EGSB reactor influent. , They are respectively Model-predicted values ​​of COD and toxic characteristic pollutants in the effluent from the EGSB reactor at specific times. For the current moment of the EGSB reactor The inlet pump flow rate is the optimal flow rate value (optimization value) searched using a control action optimization algorithm. and The removal load weights for COD and toxic characteristic pollutants are respectively. for to Total pollutant removal load within a given time period The objective function is denoted as .

[0019] In some preferred embodiments, the control action optimization algorithm sets state constraints and physical constraints.

[0020] In some preferred embodiments, the state constraints must ensure that the effluent water quality indicators (such as the concentration of free volatile fatty acids, pH, concentration of toxic characteristic pollutants, alkalinity, conductivity, etc.) all meet their threshold conditions, which can be specifically expressed as follows: In the formula, for Concentration of free VFAs in the effluent of the EGSB reactor at a given time. The corresponding dissociation constant is The maximum free VFA concentration that inhibits anaerobic microorganisms. The minimum pH suitable for the growth of anaerobic microorganisms, The minimum concentration of toxic characteristic pollutants required to inhibit the growth of anaerobic microorganisms. To meet the theoretical minimum alkalinity for EGSB reactor operation, To meet the maximum conductivity required for EGSB reactor operation.

[0021] In some preferred embodiments, physical constraints must ensure that the operating water volume of the inlet pump and return pump is lower than their maximum operating capacity, which can be specifically expressed as follows: In the formula: For the current moment of the EGSB reactor The inlet pump flow rate is the optimal flow rate value (optimization value) searched by the control action optimization algorithm. For the current moment of the EGSB reactor The return pump flow rate is the optimal flow rate value (optimization value) searched using the control action optimization algorithm. , These are the maximum operating capacities of the EGSB reactor inlet pump and return pump, respectively.

[0022] In some preferred examples, a genetic algorithm is applied to search for the operating flow rates of the inlet and outlet pumps. First, a fitness function is defined, then the population is initialized, and a roulette wheel selection algorithm is applied to calculate the probability of an individual being selected. Finally, a single-point crossover operation is performed, and the selected individuals are selected based on a set probability (…). The newly generated individuals are mutated, and further iterative updates are performed to select the individual with the highest fitness from the population as the optimal control action.

[0023] The fitness function can be specifically expressed as follows: In the formula, For the fitness function value, corresponding to to Total pollutant removal load within a given time period.

[0024] Initialized population Size is Specifically, it can be represented as follows: In the formula, The population (the set of all candidate control schemes). Population size (number of candidate control schemes). For setting the first The inlet pump flow rate and return pump flow rate are among the candidate control schemes.

[0025] The probability of an individual being selected can be calculated using the following formula. : In the formula, The fitness value for each individual can be calculated using a fitness function. The fitness value of each individual is the probability of the fitness value of all individuals.

[0026] The crossover operation is performed using a single-point crossover, and the crossover is performed with a set probability ( Mutate newly generated individuals (for example, the computer generates random numbers; if the number is less than a set probability). If the mutation occurs, then the corresponding individual is mutated; otherwise, no mutation is performed. This avoids getting trapped in local optima, and can be specifically represented as follows: In the formula, and Individuals from two different populations , Offspring formed after crossover , Represents a label. for Real-time inlet pump flow data, for Real-time return pump flow data, and These are the newly generated individuals after random mutation.

[0027] Repeat the above process again, and finally select the individual with the highest fitness as the optimal control action, that is... Optimal operating flow rates for both inlet and return pumps at all times.

[0028] In some preferred embodiments, the model predictive control-based EGSB reactor intelligent control system further includes a user interaction module; The user interaction module is used to display the digital twin model of the EGSB reactor, real-time monitoring data of the EGSB reactor sensors, output results of the EGSB reactor operation simulation and prediction model, and output results of the control action optimization algorithm.

[0029] In some preferred embodiments, the user interaction module includes an emergency dispatch module; The emergency dispatch module is used for manual start-up and shutdown of the pump.

[0030] Secondly, the present invention provides an intelligent control method for EGSB reactors based on model predictive control, which adopts the intelligent control system for EGSB reactors based on model predictive control described in the first aspect; The model predictive control-based intelligent control method for EGSB reactors includes: The data acquisition module collects water quality data of the influent and effluent of the EGSB reactor, as well as operating data of the influent pump and return pump; The data storage module aggregates multi-source data from the data acquisition module through the gateway device, performs time-series data preprocessing, and stores the data in the data warehouse. The simulation control module sends query requests to the data warehouse through the communication interface and receives returned data. It uses the EGSB reactor to run the simulation prediction model and control action optimization algorithm, generates output signals through calculation, and transmits the output signals to the inlet pump and return pump through the control foot.

[0031] Compared with the prior art, the beneficial effects of this invention are as follows: (1) Based on the optimized layout of the influent and effluent water quality sensor monitoring of the EGSB reactor, the present invention constructs an EGSB reactor operation simulation and prediction model based on the CNN-GRU-Attention deep learning algorithm, which can realize accurate prediction of the reactor effluent water quality in the future.

[0032] (2) The present invention uses a genetic algorithm to dynamically search for the optimal influent flow rate and reflux flow rate of the reactor, which can optimize the operating parameters of the EGSB reactor.

[0033] (3) The present invention sets the effluent water quality such as the concentration of toxic characteristic pollutants in the reactor, the concentration of free VFAs and pH value as constraints, and sets the removal load of organic matter and toxic characteristic pollutants as the goal, so as to maximize the performance of the reactor. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of an intelligent control system architecture for an EGSB reactor based on model predictive control according to the present invention.

[0035] Figure 2 This is a schematic diagram of the intelligent control method for an EGSB reactor based on model predictive control according to the present invention.

[0036] Figure 3 This is a comparison chart of the application performance of the present invention. Detailed Implementation

[0037] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0038] This invention constructs an edge microcontroller system based on reactor influent and effluent water quality sensing and monitoring, and big data analysis of reactor operation. It applies the CNN-GRU-Attention deep learning algorithm to build an EGSB reactor operation simulation and prediction model. Furthermore, based on model predictive control theory, it establishes reactor operation constraints and an objective function. Ensuring that the concentrations of toxic characteristic pollutants, VFAs, pH, and other effluent indicators remain within their respective threshold ranges, a genetic algorithm is applied to continuously search for the optimal influent and reflux rates, maximizing the removal load of organic matter and toxic characteristic pollutants. This invention ensures the reactor operates under optimal conditions, thereby maximizing reactor performance.

[0039] Combination Figure 1 A model predictive control-based intelligent control system for EGSB reactors is proposed. Based on the main structure of the EGSB reactor, it includes a data acquisition module, a data storage module, a simulation control module, and a user interaction module.

[0040] The data acquisition module is used to collect water quality data of the EGSB reactor's influent and effluent, as well as operating data of the feed pump and return pump. The water quality data collected by the data acquisition module comes from influent and effluent water quality monitoring sensors and water quality analyzers. The influent and effluent water quality monitoring sensors include pH and conductivity sensors, both located before the feed pump and at the EGSB reactor outlet. The water quality analyzers include a COD analyzer, an alkalinity analyzer, a volatile fatty acid analyzer, and a toxic characteristic pollutant analyzer. The volatile fatty acid analyzer's intake is only located at the EGSB reactor outlet, while the COD, alkalinity, and toxic characteristic pollutant analyzers' intakes are located before the feed pump and at the EGSB reactor outlet. Both the feed pump and return pump are variable frequency pumps. The operating data of the feed pump and return pump include pump speed, pump inlet and outlet water pressure difference, and pump power consumption. The pump flow rate is calculated based on the pump speed, pump inlet and outlet water pressure difference, and pump power consumption, and can be specifically expressed as follows: In the formula, It is the pump's flow rate. It is the pump speed. It is the pressure difference between the pump inlet and outlet. It's the pump's power consumption. , , , It is the characteristic curve value of the pump.

[0041] All data collected by the data acquisition module is transmitted via wired connection or wireless communication network.

[0042] The data storage module is used to aggregate multi-source data from the data acquisition module through the gateway device, perform time series data preprocessing, and store the data in the data warehouse.

[0043] Time series data preprocessing extracts data from the time series data based on a preset time resolution (30 min) and applies... The criteria remove outliers and simultaneously fill in the gaps using an interpolation polynomial. Specifically, this can be represented as follows: In the formula, Represents probability. Represents the mean. Represents standard deviation, Represents each original observation in the dataset. Represents the moment Estimation of missing values, Represents the smoothing coefficient, 0 < <1, Representative moment The original observation value at the previous moment, Representative moment The estimated value at the previous moment.

[0044] The analog control module is a microcontroller system used to send query requests to the data warehouse and receive returned data through the communication interface. The system's CPU (Central Processing Unit) / GPU (Graphics Processing Unit) is built-in to run the simulation prediction model and control action optimization algorithm using the EGSB reactor. It generates output signals through calculation and transmits the output signals to the inlet pump and return pump through the microcontroller control pins.

[0045] The EGSB reactor operation simulation and prediction model constructs a sub-model for each effluent water quality indicator. The inputs are time-series data including current and historical influent and effluent water quality, influent pump flow rate, and return pump flow rate. The length of the historical time series is... (60 h) is much greater than the hydraulic retention time (HRT, 12 h) of the EGSB reactor, and at the same time, at the current time The influent pump flow rate and return pump flow rate are determined by using a control action optimization algorithm to find the optimal flow rate values ​​and then applied to the operation of the EGSB reactor. ~ Within the time step; the model output is the next time step for the EGSB reactor. The water quality indicators at any given time; each sub-model is built on a CNN-GRU-Attention architecture, and the hyperparameters are tuned using a Bayesian optimization algorithm to obtain the optimal model, which can be represented as follows: In the formula, for The inflow index data matrix is ​​constantly input into the model. , , , , , , Represent to The influent COD, toxic characteristic pollutants, alkalinity, and time index of the EGSB reactor at specific times (example, time series data with a time resolution of 30 min). Values ​​can range from 1 to 48), pH, conductivity, and flow rate data; for Input the effluent recirculation index data matrix into the model at any time. , , , , , , Represent to Real-time data on COD, toxic characteristic pollutants, alkalinity, VFAs, pH, conductivity, and reflux flow rate of the EGSB reactor effluent; For model output Real-time water quality index data matrix , , , , , Represent Model-predicted values ​​of COD, toxic characteristic pollutants, alkalinity, VFAs, pH, and conductivity of the effluent from the EGSB reactor at any given time. It is an EGSB reactor operation simulation and prediction model. , , , , , These are the prediction sub-models for COD, toxic characteristic pollutants, alkalinity, VFAs, pH, and conductivity in the EGSB reactor.

[0046] The hyperparameters of each sub-model are identical, as shown in Table 1 below.

[0047] Table 1 Through training, the validation set determination coefficient (R²) of each sub-model is obtained. 2 The values ​​are all greater than 0.9.

[0048] Furthermore, the control action optimization algorithm is based on the fundamental theory of model predictive control (MPC), and sets an objective function, specifically for the EGSB reactor. to The weighted sum of the removal loads of COD and toxic characteristic pollutants within a given time period must satisfy the principle of maximizing the objective function when performing optimization, which can be specifically expressed as follows: In the formula, and They are respectively to The removal load of COD and toxic characteristic pollutants in the EGSB reactor at a given time. , They are respectively Real-time monitoring data of COD and toxic characteristic pollutants in the EGSB reactor influent. , They are respectively Model-predicted values ​​of COD and toxic characteristic pollutants in the effluent from the EGSB reactor at specific times. For the current moment of the EGSB reactor The inlet pump flow rate is the optimal flow rate value (optimization value) searched using a control action optimization algorithm. and The removal load weights for COD and toxic characteristic pollutants are respectively. for to Total pollutant removal load within a given time period The objective function is denoted as .

[0049] The aforementioned control action optimization algorithm sets state constraints and physical constraints.

[0050] The state constraints must ensure that the concentrations of free volatile fatty acids, pH, and toxic characteristic pollutants (such as chloronitrobenzene) in the effluent all meet their threshold conditions, which can be specifically expressed as follows: In the formula, for Concentration of free VFAs in the effluent of the EGSB reactor at a given time. The corresponding dissociation constant is The maximum free VFA concentration that inhibits anaerobic microorganisms. The minimum pH suitable for the growth of anaerobic microorganisms, The minimum concentration of toxic characteristic pollutants required to inhibit the growth of anaerobic microorganisms. To meet the theoretical minimum alkalinity for EGSB reactor operation, To meet the maximum conductivity required for EGSB reactor operation.

[0051] The physical constraints must ensure that the operating water volume of the inlet pump and return pump is lower than their maximum operating capacity, which can be specifically expressed as follows: In the formula: For the current moment of the EGSB reactor The inlet pump flow rate is the optimal flow rate value (optimization value) searched by the control action optimization algorithm. For the current moment of the EGSB reactor The return pump flow rate is the optimal flow rate value (optimization value) searched using the control action optimization algorithm. , These are the maximum operating capacities of the EGSB reactor inlet pump and return pump, respectively.

[0052] A genetic algorithm is applied to search for the operating flow rates of the inlet and return pumps. First, a fitness function is defined, the population is initialized, and a roulette wheel selection algorithm is used to calculate the probability of an individual being selected. Then, a single-point crossover operation is performed, and the selection is based on a set probability (…). The newly generated individuals are mutated, and further iterative updates are performed to select the individual with the highest fitness from the population as the optimal control action.

[0053] The fitness function can be specifically expressed as follows: In the formula, For the fitness function value, corresponding to to Total pollutant removal load within a given time period.

[0054] Initialized population Size is Specifically, it can be represented as follows: In the formula, The population (the set of all candidate control schemes). Population size (number of candidate control schemes). For the set number The inlet pump flow rate and return pump flow rate are among the candidate control schemes.

[0055] The probability of an individual being selected can be calculated using the following formula. : In the formula, The fitness value for each individual can be calculated using a fitness function. The fitness value of each individual is the probability of the fitness value of all individuals.

[0056] The crossover operation is performed using a single-point crossover, and the crossover is performed with a set probability ( Mutate newly generated individuals (for example, the computer generates random numbers; if the number is less than a set probability). If the mutation occurs, then the corresponding individual is mutated; otherwise, no mutation is performed. This avoids getting trapped in local optima, and can be specifically represented as follows: In the formula, and Individuals from two different populations , Offspring formed after crossover , Represents a label. for Real-time inlet pump flow data, for Real-time return pump flow data, and These are the newly generated individuals after random mutation.

[0057] Repeat the above process again, and finally select the individual with the highest fitness as the optimal control action, that is... The optimal operating flow rates for the influent and return pumps are determined at all times. The user interaction module displays the digital twin model of the EGSB reactor, real-time monitoring data from the EGSB reactor sensors, output results from the EGSB reactor operation simulation and prediction model, and output results from the control action optimization algorithm.

[0058] The user interaction module includes an emergency dispatch module. The emergency dispatch module is used for manual start-up and shutdown of the pump.

[0059] Combination Figure 2 A model predictive control-based intelligent control method for EGSB reactors is proposed, employing the aforementioned model predictive control-based intelligent control system for EGSB reactors.

[0060] The model predictive control-based intelligent control method for EGSB reactors includes: Based on the main structure of the EGSB reactor, a data acquisition module, a data storage module, a simulation control module, and a user interaction module are constructed. Data from the influent and effluent water quality monitoring sensors, water quality analyzer, influent pump, and reflux pump of the EGSB reactor are collected, pre-processed, and stored in a data warehouse. An EGSB reactor operation simulation and prediction model was constructed and placed in the simulation control module. By interacting with the data warehouse, the water quality at future times was predicted. Based on the objective of maximizing the removal load of COD and toxic characteristic pollutants in the EGSB reactor, the objective function and constraints for the intelligent control of the EGSB reactor are set. Based on model predictive control theory, a genetic algorithm is applied to dynamically search for the optimal control action to determine the optimal operating flow rate of the EGSB reactor's feed pump and return pump, and then feeds this information back to the EGSB reactor to achieve real-time control of the EGSB reactor.

[0061] like Figure 3 This is a comparison chart of the application efficiency of the present invention. After applying the intelligent control method of the present invention, the COD removal load rate of the EGSB reactor is 20.9 kgCOD / m³. 3 / d, the loading rate of chloronitrobenzene is 4.3 g / m 3 / d, compared to the traditional static control method (setting the feed water flow and return pump flow of the EGSB reactor to static values), it improved by 18.9% and 26.9% respectively.

[0062] Furthermore, it should be understood that after reading the above description of the present invention, those skilled in the art can make various alterations or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims.

Claims

1. A model predictive control-based intelligent control system for an EGSB reactor, characterized in that, Based on the main structure of the EGSB reactor, a data acquisition module, a data storage module, and a simulation control module are constructed. The data acquisition module is used to collect water quality data of the influent and effluent of the EGSB reactor, as well as operating data of the influent pump and return pump; The data storage module is used to aggregate multi-source data from the data acquisition module through the gateway device, perform time-series data preprocessing, and store the data in the data warehouse. The simulation control module is used to send query requests to the data warehouse through the communication interface and receive returned data. It uses the EGSB reactor to run the simulation prediction model and control action optimization algorithm, generates output signals through calculation, and transmits the output signals to the inlet pump and return pump through the control foot. The EGSB reactor operation simulation and prediction model constructs a sub-model for each effluent water quality indicator. The inputs are time-series data including current and historical influent and effluent water quality, influent pump flow rate, and return pump flow rate. The length of the historical time series is... It should be greater than the hydraulic retention time of the EGSB reactor, where, at the current moment... The influent pump flow rate and return pump flow rate are determined by using a control action optimization algorithm to find the optimal flow rate values ​​and then applied to the operation of the EGSB reactor. ~ Within a time step; The model output is the next time step for the EGSB reactor. The water quality indicators at any given time; each sub-model is built on the CNN-GRU-Attention architecture, and the hyperparameters are adjusted using the Bayesian optimization algorithm to obtain the optimal model; The control action optimization algorithm described above: Based on the fundamental theory of model predictive control, an objective function is set, specifically for the EGSB reactor. to The weighted sum of the removal loads of COD and toxic characteristic pollutants within a given time period must satisfy the principle of maximizing the objective function when performing optimization. Set state constraints and physical constraints; The state constraints must ensure that all effluent water quality indicators meet their threshold conditions. Physical constraints must ensure that the operating water volume of the inlet pump and return pump is lower than their maximum operating capacity.

2. The EGSB reactor intelligent control system based on model predictive control according to claim 1, characterized in that, The data acquisition module collects the influent and effluent water quality data of the EGSB reactor from the influent and effluent water quality monitoring sensors and water quality analyzers; The influent and effluent water quality monitoring sensors include pH sensors and conductivity sensors, which are installed before the influent pump and at the EGSB reactor outlet. The water quality analyzer includes a COD analyzer, an alkalinity analyzer, a volatile fatty acid analyzer, and a toxic characteristic pollutant analyzer. The volatile fatty acid analyzer's water intake is only located at the outlet of the EGSB reactor, while the COD analyzer, alkalinity analyzer, and toxic characteristic pollutant analyzer's water intakes are located before the inlet pump and at the outlet of the EGSB reactor. The operating data for the inlet and return pumps include pump speed, pump inlet and outlet water pressure difference, and pump power consumption. The pump flow rate is calculated based on the pump speed, the pressure difference between the pump inlet and outlet, and the pump power consumption.

3. The EGSB reactor intelligent control system based on model predictive control according to claim 1, characterized in that, All data collected by the data acquisition module is transmitted via wired connection or wireless communication network.

4. The EGSB reactor intelligent control system based on model predictive control according to claim 1, characterized in that, Time series data preprocessing extracts data from time series data based on a preset time resolution and applies... The criteria remove outliers and fill them with interpolation polynomials.

5. The intelligent control system for EGSB reactors based on model predictive control according to claim 1, characterized in that, The analog control module is a single-chip microcomputer system; Both the inlet pump and the return pump are variable frequency pumps.

6. The EGSB reactor intelligent control system based on model predictive control according to claim 1, characterized in that, A genetic algorithm is used to search for the operating flow rates of the inlet and return pumps. First, a fitness function is defined, and the population is initialized. The roulette wheel selection method is used to perform the selection operation, and the probability of an individual being selected is calculated. Then, a single-point crossover operation is performed, and the newly generated individuals are mutated according to a set probability. The process is iteratively updated, and the individual with the highest fitness is selected from the population as the optimal control action.

7. The EGSB reactor intelligent control system based on model predictive control according to claim 1 or 2, characterized in that, The model predictive control-based EGSB reactor intelligent control system also includes a user interaction module; The user interaction module is used to display the digital twin model of the EGSB reactor, real-time monitoring data of the EGSB reactor sensors, output results of the EGSB reactor operation simulation and prediction model, and output results of the control action optimization algorithm. The user interaction module includes an emergency dispatch module; The emergency dispatch module is used for manual start-up and shutdown of the pump.

8. A method for intelligent control of an EGSB reactor based on model predictive control, characterized in that, The EGSB reactor intelligent control system based on model predictive control as described in any one of claims 1-7 is adopted; The model predictive control-based intelligent control method for EGSB reactors includes: The data acquisition module collects water quality data of the influent and effluent of the EGSB reactor, as well as operating data of the influent pump and return pump; The data storage module aggregates multi-source data from the data acquisition module through the gateway device, performs time-series data preprocessing, and stores the data in the data warehouse. The simulation control module sends query requests to the data warehouse through the communication interface and receives returned data. It uses the EGSB reactor to run the simulation prediction model and control action optimization algorithm, generates output signals through calculation, and transmits the output signals to the inlet pump and return pump through the control foot.

Images