UASB reactor layered adaptive regulation and control system, method and device based on machine learning, electronic equipment and storage medium

Abstract

The invention relates to a machine learning-based UASB (Upflow Anaerobic Sludge Blanket) reactor layered self-adaptive regulation and control system, which comprises a UASB reactor, a monitoring unit, an execution unit and a control unit. The invention further relates to a UASB reactor layered self-adaptive regulation and control method based on machine learning. The method comprises the steps of collecting data, predicting an evolution trend, generating a control instruction, sending the control instruction, and comparing the evolution trend with the control instruction. The invention further relates to a UASB reactor layering self-adaptive regulation and control device based on machine learning. The UASB reactor layering self-adaptive regulation and control device comprises an acquisition module, an import module, a generation module, a sending module and a comparison module. The invention further relates to an electronic device and a storage medium for UASB reactor layered self-adaptive regulation and control based on machine learning. According to the UASB reactor layered self-adaptive regulation and control system, method and device based on machine learning, the electronic equipment and the storage medium, the problems that in the prior art, the overall potential of a UASB reactor is limited, the process is unbalanced, and dynamic changes of inflow water quality and loads are difficult to cope with can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of biological wastewater treatment, and particularly to a hierarchical adaptive regulation system, method, device, electronic device and storage medium of a UASB reactor based on machine learning. Background Art

[0002] Upflow anaerobic sludge bed (UASB) reactors are widely used due to their high efficiency in treating high-concentration organic wastewater. However, their stable and efficient operation faces inherent challenges. Anaerobic microbial communities (especially methanogenic archaea) are extremely sensitive to environmental conditions (such as temperature, pH, redox potential ORP), and have a long generation time, making the system startup, commissioning, and recovery process after being impacted long and difficult. More importantly, anaerobic digestion is a complex multi-stage biochemical process, involving main steps such as hydrolysis, acidogenesis, acetogenesis, and methanogenesis, which are completed collaboratively by microbial flora with different functions. The optimal growth and metabolic environmental conditions of these different functional flora (for example, the ideal ORP range, suitable pH value, specific carbon-nitrogen-phosphorus C:N:P nutrient ratio, volatile fatty acid VFA concentration threshold, etc.) often have significant differences or even conflicts.

[0003] Traditional UASB reactor design and control strategies usually regard the reactor as a whole or conduct rough segmentation. Their monitoring points are limited, and the control means are single (such as overall adjustment of influent flow rate, temperature, or unified addition of pH / alkalinity regulators), making it difficult to meet the specific and differentiated requirements of the microbial communities with different dominant functions naturally formed in different vertical regions of the reactor for the microenvironment. This "one-size-fits-all" or passive regulation method based on overall state feedback often results in operating conditions being a compromise, unable to make the microorganisms in any functional area be in their absolutely optimal metabolic state, thus limiting the overall potential of the reactor, easily causing process imbalance (such as acid accumulation, ORP fluctuation), and being difficult to cope with the dynamic changes of influent water quality and load.

[0004] The prior art generally lacks a system and method that can actively and precisely create and maintain differentiated and individually optimized local microenvironments (especially the key ORP and C:N:P ratio) for the specific biochemical requirements of different functional levels (such as hydrolysis acidification layer, methanogenesis layer) inside the UASB reactor. There is an urgent need for a more intelligent and adaptable regulation strategy that can go beyond simple passive responses and achieve active and precise management of the complex microecology inside the UASB.

[0005] Therefore, the prior art has defects and needs to be improved and developed. Summary of the Invention

[0006] An embodiment of the invention provides a hierarchical adaptive regulation method, device, electronic device, and storage medium for a UASB reactor based on machine learning, which is used to solve the problems in the prior art that the UASB reactor lacks active and precise microecological management, resulting in the limitation of the overall potential of the UASB reactor, causing process imbalance, and being difficult to cope with the dynamic changes of influent water quality and load.

[0007] In the first aspect of the embodiment of the present invention, a hierarchical adaptive regulation system for a UASB reactor based on machine learning is provided, including:

[0008] A UASB reactor, which is divided into at least two functional regions with different functions in the vertical direction according to the main functions of microorganisms;

[0009] A monitoring unit, which includes a number of sensors, and several of the sensors for real-time and in-situ capturing of microenvironment data are arranged in each functional region;

[0010] An execution unit, which includes an influent distribution network and a chemical dosing system; the influent distribution network is configured to independently adjust the influent flow rate into each functional region; the chemical dosing system is provided with a number of chemical sources and dosing devices adapted to the chemical sources, and the chemical dosing system is configured to independently dose the chemicals of the chemical sources into any of the functional regions by using the dosing devices;

[0011] A control unit, which is electrically connected to the monitoring unit and the execution unit, and at least one trained machine learning model is embedded in the control unit. The control unit is configured to receive the microenvironment data of each functional region in real time to predict the evolution trend of the future state of each functional region based on the machine learning model, and calculate a control strategy according to the target operating state of each functional region to send control instructions for regulating the influent flow rate and chemical dosing dose of each functional region to the execution unit.

[0012] Further, the functional regions at least include a region mainly for hydrolysis and acidification and a region mainly for methane production from bottom to top.

[0013] Further, the sensors at least include one of an oxidation-reduction potential sensor for accurately characterizing the oxidation-reduction state of the functional region, a sensor for monitoring relevant parameters reflecting the local carbon-nitrogen-phosphorus nutrient balance status of the functional region, a temperature sensor, a dissolved methane sensor, a hydrogen partial pressure sensor, and a sensor for detecting the pH value.

[0014] Further, the agent source at least includes one of a carbon source for adjusting the local carbon-nitrogen-phosphorus ratio, a nitrogen source for adjusting the local carbon-nitrogen-phosphorus ratio, a phosphorus source for adjusting the local carbon-nitrogen-phosphorus ratio, a trace oxidant for adjusting the local redox potential, a reducing agent for adjusting the local redox potential, an electron shuttle for adjusting the local redox potential, a conductive particle for adjusting the local redox potential, an acidic reagent for adjusting the local pH, and a basic reagent for adjusting the local pH.

[0015] Further, the machine learning model at least includes one of a reinforcement learning agent model, a recurrent neural network model, a graph neural network model, a support vector machine model, and a hybrid model.

[0016] Further, the evolution trend at least includes one of the accumulation rate of volatile fatty acids, the downward trend of the redox potential, the change in pH value, and the change in the mass and ratio of carbon-nitrogen-phosphorus nutrients.

[0017] In the second aspect of the embodiments of the present invention, a hierarchical adaptive regulation method for a UASB reactor based on machine learning is provided. The hierarchical adaptive regulation method for a UASB reactor based on machine learning is applied to the hierarchical adaptive regulation system for a UASB reactor based on machine learning, and includes:

[0018] Real-time collect the microenvironment data monitored by sensors arranged in each functional area of the UASB reactor;

[0019] Import the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future state of each functional area;

[0020] Based on the evolution trend, generate control instructions for each functional area including the inlet water flow distribution and the agent dosing amount;

[0021] Send the control instructions to the execution unit to distribute the inlet water flow and add agents to each functional area through the inlet water distribution network and the agent dosing system respectively;

[0022] Compare the evolution trend and the control instructions to online adjust, update parameters, and iteratively optimize the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

[0023] In the third aspect of the embodiments of the present invention, a hierarchical adaptive regulation device for a UASB reactor based on machine learning is provided, including:

[0024] An acquisition module for real-time collecting the microenvironment data monitored by sensors arranged in each functional area of the UASB reactor;

[0025] An import module for importing the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future state of each of the functional regions;

[0026] A generation module for generating control instructions for each of the functional regions, including influent flow distribution and chemical dosing dosage, based on the evolution trend;

[0027] A sending module for sending the control instructions to an execution unit to distribute influent flow and add chemicals to each of the functional regions through an influent distribution network and a chemical dosing system respectively;

[0028] A comparison module for comparing the evolution trend and the control instructions to online adjust, update parameters and iteratively optimize the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

[0029] In a fourth aspect of an embodiment of the present invention, there is provided an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the hierarchical adaptive regulation method of a UASB reactor based on machine learning is implemented.

[0030] In a fifth aspect of an embodiment of the present invention, there is provided a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the hierarchical adaptive regulation method of a UASB reactor based on machine learning is implemented.

[0031] Advantageous effects:

[0032] As can be seen from the above technical solutions, the present invention provides a hierarchical adaptive regulation system, method, device, electronic device and storage medium of a UASB reactor based on machine learning:

[0033] 1. Maximize synergistic efficiency: Simultaneously meet the specific harsh environmental requirements (especially ORP and C:N:P) of different functional microbial communities in their respective advantageous regions, enabling each link such as hydrolysis, acid production, and methane production to operate at a rate close to its theoretical optimum, improving the overall organic matter removal efficiency, methane production rate, and energy recovery efficiency.

[0034] 2. Enhance system robustness and resilience: The machine learning model (ML) can identify potential risks (such as acidification and toxic shock) in advance through prediction and perform precise and regionalized preventive interventions, greatly enhancing the system's ability to resist load and environmental fluctuations and shortening the recovery time after being disturbed. Through the dual mechanisms of prediction and real-time regulation, it can quickly respond and formulate reasonable coping strategies in the case of load mutation, water quality disturbance, etc., significantly improving the operation reliability and reducing the risk of over-standard emissions.

[0035] 3. Achieve precise and efficient utilization of pharmaceuticals: Instead of blindly dosing pharmaceuticals or dosing based on overall indicators, they are precisely delivered according to the actual needs of specific areas, reducing the total pharmaceutical consumption and operating costs. Setting up multiple types of pharmaceutical sources and precisely controlling their dosing concentrations and positions helps maintain a stable micro-ecological environment in each area of the reactor, improve the local nutrient structure or electron acceptor supply conditions, thereby stabilizing the microbial activity and reducing the probability of instability phenomena such as volatile fatty acid accumulation and methanation delay.

[0036] 4. Facilitate in-depth understanding of complex anaerobic processes: The continuous operation and the learning process of the ML model can continuously adjust, update, and iterate the ML model, thereby revealing more complex mechanisms of the micro-ecological interactions inside the UASB reactor.

[0037] 5. Achieve differential and precise regulation of functional areas inside the UASB reactor: By vertically partitioning the reactor according to microbial functions and combining the sensors and independent control devices arranged in each area, each functional area can separately obtain operation data and implement directional control, breaking through the adaptation bottleneck of the traditional integrated regulation strategy to system heterogeneity and improving the granularity and adaptability of system regulation.

[0038] 6. Enhance the prediction ability of the evolution trend of complex systems: By introducing machine learning models with the ability to model time series characteristics (such as recurrent neural networks, graph neural networks, reinforcement learning agents, etc.), historical and real-time data can be used to deduce future evolution trends, identify possible fluctuations, imbalances, or faults that may occur during system operation in advance, and significantly improve the feedforward regulation ability of system operation. Combining a feedback mechanism to online train the control model and update parameters can continuously optimize the regulation strategy under conditions of operating fluctuations or water quality changes, improve the stability and adaptive ability of the system during long-term operation, and meet the intelligent control requirements in complex application scenarios.

[0039] It should be understood that all combinations of the foregoing concepts and additional concepts described in more detail below can be regarded as part of the inventive subject matter of the present disclosure as long as such concepts do not conflict with each other.

[0040] The foregoing and other aspects, embodiments, and features of the teachings of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of exemplary embodiments, will be apparent from the following description or learned through the practice of specific embodiments according to the teachings of the present invention. Description of the Drawings

[0041] The accompanying drawings are not drawn to scale with respect to actual reference objects. In the accompanying drawings, each identical or approximately identical component shown in each figure may be denoted by the same reference numeral. For the sake of clarity, not every component is labeled in each figure. Now, embodiments of various aspects of the present invention will be described by way of example and with reference to the accompanying drawings, wherein:

[0042] Figure 1 This is a flowchart of the hierarchical adaptive regulation method for a UASB reactor based on machine learning in the embodiments of the present application. Detailed implementation manners

[0043] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings understood by those of ordinary skill in the art in the field to which the present invention pertains.

[0044] The terms "first", "second", and similar terms used in the specification and claims of this patent application for the present invention do not denote any order, quantity, or importance, but are only used to distinguish different components. Similarly, unless the context clearly indicates otherwise, singular forms such as "a", "an", or "the" and similar terms do not denote a limitation in quantity, but rather indicate the presence of at least one. The terms "including" or "comprising" and similar terms are intended to mean that the elements or items appearing before "including" or "comprising" cover the features, wholes, steps, operations, elements, and / or components listed after "including" or "comprising", and do not exclude the existence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations. The terms "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0045] In the prior art, the UASB reactor lacks active and precise microecological management, the overall potential of the USAB reactor is limited, resulting in process imbalance and making it difficult to cope with the dynamic changes in influent water quality and load.

[0046] In view of this, the embodiments of the present invention provide a hierarchical adaptive regulation system for a UASB reactor based on machine learning, including:

[0047] UASB reactor, the UASB reactor is divided into at least two functional areas with different functions in the vertical direction according to the main functions of microorganisms.

[0048] The functional areas with different functions are divided in the vertical direction according to the main functions of microorganisms. The number of functional areas can be greater than 2, as long as the functional areas that can simulate the spatial division characteristics of the real biological community can be divided and set in the vertical direction. The main function of microorganisms refers to the main biochemical reactions carried out by microorganisms in the real biological community.

[0049] Monitoring unit, the monitoring unit includes a number of sensors, and several sensors for capturing microenvironment data in real time and in situ are arranged in each functional area.

[0050] At key positions inside each functional area, multiple sensors are densely or strategically arranged to capture the microenvironment data of this specific area in real time and in situ. By collecting parameters such as redox potential, temperature, nutrient ratio, pH value, dissolved gas concentration, etc. in real time, the real-time perception of the microecological environment is realized.

[0051] Execution unit, the execution unit includes an influent distribution network and a chemical dosing system; the influent distribution network is configured to independently adjust the influent flow rate into each functional area; the chemical dosing system is provided with a number of chemical sources and dosing devices adapted to the chemical sources. The dosing device uses the device for dosing chemicals in the existing technology. The chemical dosing system is configured to independently dose the chemicals of the chemical source into any functional area by using the dosing device.

[0052] The execution unit is used for water distribution and chemical dosing, including a finely adjustable influent distribution network and a multi-channel and multi-point chemical dosing system. The influent distribution network is designed to be able to independently and accurately control the influent flow rate ratio or absolute flow rate distributed to each of the functional areas according to the control instructions, allowing different hydraulic loads or dilution degrees to be applied to different areas. The chemical dosing system includes multiple independent chemical sources and dosing devices, and can independently, accurately and on-demand dose a variety of chemical agents into each of the functional areas according to the control instructions, or pre-mix them with the influent flow of the area before entering the area. In addition, the physical structures of the stratified influent distribution network and the chemical dosing system can be further optimized through existing public technologies, such as AI design methods combining computational fluid dynamics simulation and generative adversarial network, to ensure that the control instructions can be efficiently and evenly executed to the target area.

[0053] A control unit, electrically connected to a monitoring unit and an execution unit. At least one trained machine learning model is embedded in the control unit. The control unit is configured to receive in real time the microenvironment data of each functional area to predict the evolution trend of the future state of each functional area based on the machine learning model and calculate a control strategy according to the target operating state of each functional area to send a control instruction for regulating the influent flow rate and chemical dosing dosage of each functional area to the execution unit.

[0054] As the brain of the machine learning-based hierarchical adaptive regulation system of the UASB reactor, the control unit is electrically connected to the monitoring unit and the execution unit. The core of the control unit is one or more machine learning models, etc. The control unit is configured to: receive and integrate in real time the high-dimensional, multi-parameter sensor data streams from all functional areas. Use the ML model to deeply learn and understand the complex, non-linear, time-varying biochemical kinetics and environmental response relationships within and between each functional area. Actively predict the future state evolution trend of each area, for example, predict the cumulative rate of VFA, the downward trend of ORP, the limitation of specific nutrients, etc. Based on the prediction results and the preset regional optimization objectives, use the ML model or combined with optimization algorithms to calculate the best control strategy combination that can actively guide and maintain each area at its respective optimal microenvironment state. This strategy includes the influent allocation amount for each area and the specific types and precise dosages of various chemicals added to each area, generate and send precise control instructions to the hierarchical execution unit to achieve differential, forward-looking, and adaptive regulation of the microenvironment of each area. The regional optimization objectives refer to the ideal ORP range, C:N:P ratio, pH range, etc. customized for each area. The design of the independent execution unit avoids the cross-regional diffusion of chemicals, which conforms to the theory of microbial niche separation.

[0055] The machine learning-based hierarchical adaptive regulation system for UASB reactors does not merely passively regulate the overall state of the reactor. Instead, it obtains real-time state information of each functional area through hierarchical monitoring, and uses machine learning algorithms for intelligent analysis, prediction, and decision-making. Then, through hierarchical water distribution and chemical dosing for precise execution, within each key functional area inside the UASB reactor, it actively and continuously creates and maintains a differentiated and optimized local microenvironment that is conducive to the efficient functioning of the dominant microbial community in that area, especially optimizing the ORP and C:N:P ratio. By establishing a hierarchical control system for UASB reactors driven by machine learning models, precise independent control of the influent and chemical dosing in different functional areas is achieved, enhancing the dynamic response ability and operational stability of the reactor. This system improves the adaptability of the reactor under complex organic load changes through a monitoring-prediction-control closed-loop structure. Based on the natural distribution principle of the microbial function stratification inside the UASB reactor, multi-dimensional parameters are obtained by setting independent sensors and actuators in different areas, and the machine learning model embedded is used to non-linearly model the microecological evolution trend. Combining the prediction results, regionalized regulation instructions are generated, thus realizing feedforward active regulation and avoiding relying on lagging feedback responses.

[0056] In some embodiments, the functional areas at least include a hydrolysis-acidification-dominated area and a methane-production-dominated area from bottom to top. In a UASB reactor, the anaerobic microbial population forms a metabolic gradient in the vertical direction: mostly hydrolysis-acidification bacteria at the bottom and mostly methane-producing bacteria at the top. This distribution results in different optimal requirements for environmental conditions such as nutrients, redox state, and pH in each area. Therefore, zoning regulation of the reactor can enable each type of microorganism to operate under its optimal conditions, improving the overall reaction efficiency.

[0057] In some embodiments, the sensors at least include one of an oxidation-reduction potential sensor for accurately characterizing the redox state of the functional area, a sensor for monitoring relevant parameters reflecting the local carbon-nitrogen-phosphorus nutrient balance status of the functional area, a temperature sensor, a dissolved methane sensor, a hydrogen partial pressure sensor, and a sensor for detecting the pH value.

[0058] A multi-type microenvironment sensor network is constructed, which can real-time monitor key parameters such as the redox state, temperature, nutrient ratio, dissolved gas concentration, and pH value in different functional areas, providing high-dimensional input data for the machine learning model and improving the accuracy and adaptability of model prediction. The redox potential sensor is used to evaluate the balance of electron donors and acceptors in the area, reflecting the microbial metabolic activity; the temperature sensor is used to regulate the reaction kinetics; the dissolved methane and hydrogen partial pressure sensors monitor the methane production efficiency and intermediate metabolites respectively; the carbon, nitrogen, and phosphorus nutrient parameter sensors reflect the limiting factors for microbial growth; the pH sensor is used to control the accumulation of acidification products and the survival window of methanogens. These in-situ sensors combine to form a multi-parameter monitoring system for the microenvironment, providing a data basis for subsequent prediction and control.

[0059] In some embodiments, the reagent source at least includes one of a carbon source for regulating the local carbon, nitrogen, and phosphorus ratio, a nitrogen source for regulating the local carbon, nitrogen, and phosphorus ratio, a phosphorus source for regulating the local carbon, nitrogen, and phosphorus ratio, a trace oxidant for regulating the local redox potential, a reducing agent for regulating the local redox potential, an electron shuttle for regulating the local redox potential, a conductive particle for regulating the local redox potential, an acidic reagent for regulating the local pH, and a basic reagent for regulating the local pH.

[0060] By providing various reagent sources covering redox regulation, nutrient regulation, pH regulation, etc. in the system, it provides a means for differential regulation of different functional areas, enhancing the system's ability to repair and stabilize the imbalanced microecology. The carbon / nitrogen / phosphorus source is used to regulate the nutrient ratio to make C:N:P tend to an appropriate ratio to maintain microbial activity; the oxidant and reducing agent change the local ORP value to control the metabolic activity of different bacterial groups; the electron shuttle and conductive particle improve the efficiency of anaerobic reactions by promoting the electron transfer rate; the acid / base agent regulates the local pH to maintain neutral to weakly alkaline to ensure the activity of methanogens. The above-mentioned reagents can be independently distributed to the specified functional areas through the cooperation of the control unit and the execution unit to achieve refined chemical regulation.

[0061] In some embodiments, the machine learning model at least includes one of a reinforcement learning agent model, a recurrent neural network model, a graph neural network model, a support vector machine model, and a hybrid model.

[0062] By adapting to different and multiple machine learning models, the applicability and scalability of the system are enhanced, and the optimal modeling method can be selected according to the characteristics of the target variable, data type, and control requirements. The reinforcement learning agent model is suitable for policy learning, and the optimal control policy is updated through the reward mechanism; the recurrent neural network model is suitable for predicting time series data, such as temperature changes and ORP trends; the graph neural network can process data between regions with spatial correlations; the support vector machine is suitable for classification and regression tasks under medium and small-scale samples; the hybrid model combines the advantages of the above models to improve the prediction accuracy and generalization ability. Modeling the dynamic evolution of the UASB microecosystem through the above algorithms helps to optimize control response and time scheduling.

[0063] In some embodiments, the evolution trend at least includes one of the cumulative rate of volatile fatty acids, the downward trend of redox potential, the change in pH value, and the change in the mass and ratio of carbon, nitrogen, and phosphorus nutrients.

[0064] By defining the predicted output content of the machine learning model, it helps to achieve future-oriented feedforward control and can construct a control instruction generation strategy based on trend changes. The cumulative rate of volatile fatty acids reflects whether there is excessive accumulation in the acidification stage; the downward trend of redox potential indicates an enhanced anaerobic state of the system, which may inhibit some microorganisms; the change in pH value is a feedback signal for key reaction processes; the change in the ratio of C:N:P determines whether the microbial metabolism is restricted. Through trend prediction, the purpose of early intervention in the system and preventing the deterioration of the operation of functional areas can be achieved.

[0065] Another embodiment of the present invention also provides a hierarchical adaptive regulation method for a UASB reactor based on machine learning. The hierarchical adaptive regulation method for a UASB reactor based on machine learning is applied to a hierarchical adaptive regulation system for a UASB reactor based on machine learning, referring to Figure 1 , including:

[0066] Step S102: Real-time collect the microenvironment data monitored by sensors set in each functional area of the UASB reactor.

[0067] Through the hierarchical fine monitoring unit, multi-dimensional microenvironment data such as ORP, C:N:P related parameters, pH, and VFA in each functional area of the UASB reactor are collected in real time and at high frequency.

[0068] Step S104: Import the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future state of each functional area.

[0069] The control unit processes the received multi-region data stream using the trained machine learning model, not only analyzing the deviation between the current state and the independent optimization objectives of each region, but more importantly, predicting the evolution trend of the key parameters in each region over a period of time in the future, and evaluating the impacts of different potential regulation actions, such as changing the influent distribution of a certain region, adding specific chemicals to a certain region, etc., on the future states of each region and the overall system performance.

[0070] Step S106: Based on the evolution trend, generate control instructions for each functional region, including influent flow distribution and chemical dosage.

[0071] Step S108: Send the control instructions to the execution unit to distribute the influent flow and add chemicals to each functional region through the influent distribution network and the chemical dosing system respectively.

[0072] Based on the predictions of the ML model and the optimization calculation results, the control unit generates a set of differentiated and optimized control instructions for all functional regions. These instructions are converted by the execution unit into the precise distribution of the influent flow for each region and the precise control of the types and dosages of various chemicals added to each region. The goal of the regulation is to actively prevent the occurrence of deviation from the optimal state, or to quickly and accurately pull the environment of the region back to its specific optimal range when the deviation occurs.

[0073] Step S110: Compare the evolution trend and the control instructions to perform online adjustment, parameter update, and iterative optimization of the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

[0074] During the operation of the system, data is continuously collected. The ML model performs online learning, parameter update, or model iteration according to the actual operation effect, continuously optimizing its understanding of the system dynamics and the effectiveness of the control strategy, and achieving long-term adaptability to changing working conditions, such as changes in influent characteristics and microbial community succession.

[0075] By providing a machine learning-based UASB reactor control method, covering the entire process from data acquisition, model inference, strategy generation, execution feedback to model iteration, ensuring the closed-loop operation of the control process, and realizing the self-adaptive and self-optimizing capabilities of the reactor. Taking the time-series data of the UASB reactor operation state as the input, calculating the evolution trend through the trained prediction model, and generating the control quantity based on the target state in the strategy module. The control quantity is input into the actuator to complete the adjustment of water flow and chemicals; after execution, continue to collect feedback data for correcting the model parameters, thus forming a "perception - decision - execution - feedback" closed-loop reinforcement control system, making the system operation more precise, efficient, and stable.

[0076] Another embodiment of the present invention further provides a hierarchical adaptive regulation device for a UASB reactor based on machine learning, including:

[0077] A collection module, configured to collect in real time the microenvironment data monitored by sensors disposed in each functional area of the UASB reactor;

[0078] An import module, configured to import the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future states of each functional area;

[0079] A generation module, configured to generate control instructions for each functional area, including influent flow distribution and chemical dosing dosage, based on the evolution trend;

[0080] A sending module, configured to send the control instructions to an execution unit to distribute the influent flow and add chemicals to each functional area through an influent distribution network and a chemical dosing system respectively;

[0081] A comparison module, configured to compare the evolution trend and the control instructions to perform online adjustment, parameter update and iterative optimization of the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

[0082] Another embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements a hierarchical adaptive regulation method for a UASB reactor based on machine learning.

[0083] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc. The processor is the control center of the data processing device of the gateway, and connects various parts of the data processing device of the entire gateway through various interfaces and lines.

[0084] Another embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements a hierarchical adaptive regulation method for a UASB reactor based on machine learning.

[0085] The memory may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the processor and the like. In addition, the memory is preferably but not limited to a high-speed random access memory. For example, it can also be a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory may also optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0086] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiment methods, a program can be used to instruct relevant hardware to complete. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memories.

[0087] In summary, the hierarchical adaptive regulation system, method, device, electronic equipment and storage medium based on machine learning provided by the present invention change the control concept of traditional UASB reactors, shifting from passively responding to the overall state to actively shaping and maintaining the differentiated optimal microenvironments of each functional area. Through precise, machine learning-based intelligent regulation, the following are achieved: Maximize synergy: Simultaneously meet the specific and demanding environmental requirements (especially ORP and C:N:P) of different functional microbial communities in their respective advantageous regions, enabling each link such as hydrolysis, acidogenesis, and methanogenesis to operate at a rate close to its theoretical optimum, improving the overall organic matter removal efficiency, methane production rate, and energy recovery efficiency. Enhance system robustness and resilience: The machine learning (ML) model can identify potential risks (such as acidification and toxicant shock) in advance through prediction and perform precise and regionalized preventive interventions, greatly enhancing the system's ability to resist load and environmental fluctuations and shortening the recovery time after being perturbed. Through the dual mechanisms of prediction and real-time regulation, it can quickly respond and formulate reasonable coping strategies in the case of sudden load changes, water quality disturbances, etc., significantly improving the operation reliability and reducing the risk of over-standard emissions. Achieve precise and efficient utilization of chemicals: Chemicals are no longer added blindly or based on overall indicators, but are precisely delivered according to the actual needs of specific regions, reducing the total chemical consumption and operating costs. Setting multiple types of chemical sources and finely controlling their dosing concentrations and positions helps maintain the stable microecological environment in each region of the reactor, improve the local nutritional structure or electron acceptor supply conditions, thereby stabilizing microbial activity and reducing the probability of instability phenomena such as the accumulation of volatile fatty acids and methane production delay. Facilitate in-depth understanding of complex anaerobic processes: The continuous operation and learning process of the ML model can continuously adjust, update, and iterate the ML model, thereby revealing more complex mechanisms of the internal microecological interactions in UASB reactors. Achieve differentiated and precise regulation of functional areas in UASB reactors: By vertically partitioning the reactor according to microbial functions and combining the sensors and independent regulation devices arranged in each area, each functional area can respectively obtain operation data and implement directional control, breaking through the adaptation bottleneck of traditional integrated regulation strategies to system heterogeneity and improving the granularity and adaptability of system regulation. Enhance the prediction ability of the evolution trend of complex systems: By introducing machine learning models with the ability to model temporal characteristics (such as recurrent neural networks, graph neural networks, reinforcement learning agents, etc.), the future evolution trend can be deduced using historical and real-time data, and fluctuations, imbalances, or failures that may occur during system operation can be identified in advance, significantly enhancing the feedforward regulation ability of system operation. Combining the feedback mechanism to conduct online training and parameter update of the control model can continuously optimize the regulation strategy under the conditions of operating condition fluctuations or water quality changes, improving the long-term operation stability and adaptive ability of the system and meeting the intelligent control requirements in complex application scenarios.

[0088] Although the present invention has been disclosed above in preferred embodiments, it is not intended to limit the present invention. Those of ordinary skill in the technical field to which the present invention pertains can make various modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to what is defined by the claims.

Claims

1. A hierarchical adaptive control system for UASB reactors based on machine learning, characterized in that Comprising: A UASB reactor, which is divided into at least two functional regions with different functions in the vertical direction according to the main functions of microorganisms; A monitoring unit, which includes a number of sensors, and a number of the sensors for capturing microenvironment data in real time and in-situ are arranged in each functional region; An execution unit, which includes an influent distribution network and a chemical dosing system; the influent distribution network is configured to independently adjust the influent flow rate entering each functional region; the chemical dosing system is provided with a number of chemical sources and dosing devices adapted to the chemical sources, and the chemical dosing system is configured to independently dose the chemicals of the chemical sources into any of the functional regions by using the dosing devices; A control unit, which is electrically connected to the monitoring unit and the execution unit, and at least one trained machine learning model is embedded in the control unit. The control unit is configured to receive the microenvironment data of each functional region in real time to predict the evolution trend of the future state of each functional region based on the machine learning model, and calculate a control strategy according to the target operating state of each functional region to send control instructions for regulating the influent flow rate and chemical dosing dose of each functional region to the execution unit.

2. The hierarchical adaptive regulation system of a UASB reactor based on machine learning according to claim 1, wherein The functional region at least includes a region mainly for hydrolysis and acidification and a region mainly for methane production from bottom to top.

3. The hierarchical adaptive regulation system of a UASB reactor based on machine learning according to claim 1, wherein The sensor at least includes an oxidation-reduction potential sensor for accurately characterizing the oxidation-reduction state of the functional region, a sensor for monitoring relevant parameters reflecting the local carbon, nitrogen and phosphorus nutrient balance status of the functional region, a temperature sensor, a dissolved methane sensor, a hydrogen partial pressure sensor, and a sensor for detecting the pH value.

4. A hierarchical adaptive regulation system for a UASB reactor based on machine learning according to claim 1, characterized in that, The chemical source at least includes a carbon source for adjusting the local carbon, nitrogen and phosphorus ratio, a nitrogen source for adjusting the local carbon, nitrogen and phosphorus ratio, a phosphorus source for adjusting the local carbon, nitrogen and phosphorus ratio, a trace oxidant for adjusting the local oxidation-reduction potential, a reducing agent for adjusting the local oxidation-reduction potential, an electron shuttle for adjusting the local oxidation-reduction potential, a conductive particle for adjusting the local oxidation-reduction potential, an acidic reagent for adjusting the local pH, and a basic reagent for adjusting the local pH.

5. The hierarchical adaptive regulation system of a UASB reactor based on machine learning according to claim 1, characterized in that, The machine learning model at least includes a reinforcement learning agent model, a recurrent neural network model, a graph neural network model, a support vector machine model, and a hybrid model.

6. The hierarchical adaptive regulation system of a UASB reactor based on machine learning according to claim 1, characterized in that The evolution trend at least includes one of the cumulative rate of volatile fatty acids, the downward trend of oxidation-reduction potential, the change of pH value, and the change of the mass and ratio of carbon, nitrogen and phosphorus nutrients.

7. A hierarchical adaptive regulation method for UASB reactors based on machine learning, characterized in that, The method for hierarchical adaptive regulation of a UASB reactor based on machine learning is applied to the hierarchical adaptive regulation system of a UASB reactor based on machine learning as claimed in claim 1, and includes: Real-time collecting the microenvironment data monitored by the sensors arranged in each functional region of the UASB reactor; Import the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future state of each of the functional regions; Generate control instructions for each of the functional regions, including influent flow distribution and chemical dosing dosage, based on the evolution trend; Send the control instructions to the execution unit to distribute influent flow and add chemicals to each of the functional regions through the influent distribution network and the chemical dosing system respectively; Compare the evolution trend and the control instructions to perform online adjustment, parameter update, and iterative optimization of the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

8. A hierarchical adaptive control device for UASB reactors based on machine learning, characterized in that, Comprising: A collection module for collecting in real time the microenvironment data monitored by sensors arranged in each functional region of the UASB reactor; An import module for importing the microenvironment data into a control unit configured with at least one trained machine learning model to predict the evolution trend of the future state of each of the functional regions; A generation module for generating control instructions for each of the functional regions, including influent flow distribution and chemical dosing dosage, based on the evolution trend; A sending module for sending the control instructions to the execution unit to distribute influent flow and add chemicals to each of the functional regions through the influent distribution network and the chemical dosing system respectively; A comparison module for comparing the evolution trend and the control instructions to perform online adjustment, parameter update, and iterative optimization of the machine learning model according to the deviation between the operation effect of the control instructions and the evolution trend.

9. An electronic device, characterized in that, Comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, where when the computer program is executed by the processor, it implements the hierarchical adaptive regulation method of the UASB reactor based on machine learning as claimed in claim 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the hierarchical adaptive regulation method of the UASB reactor based on machine learning as claimed in claim 7.

Images