6 /6
Blower 1 Blower 2 Blower 3 Temperature, ammonia nitrogen, COD, TKN Blower room (PSO-BP neural -- -- --- -- -..- --. -- -- -- --- - -- -- -- -- - -- -- -- - --- - -- network-based adaptive control)
Aerobic tank 1# Aerobic tank 2# Aerobic tank 3#
FIG. 9
ADAPTIVE CONTROL METHOD AND SYSTEM FOR AERATION PROCESS TECHNICAL FIELD The present disclosure relates to the field of wastewater treatment, in particular to an adaptive control method and system for an aeration process. BACKGROUND The wastewater treatment industry is an energy-intensive industry (ElI), which mainly consumes chemicals, electrical energy and combustion heat. The electricity consumption of an urban wastewater treatment plant accounts for about 60-90% of the direct energy consumption of the whole plant. As the core of wastewater treatment, the biological treatment system undertakes the task of removing main pollutants in the wastewater. The energy consumption required for operating the biological treatment system accounts for about 50-70% of the energy consumption of the whole plant. The optimization and transformation of the aeration process of the biological treatment system of the wastewater treatment plant to improve its stability and utilization is the key to the wastewater treatment plant's discharge compliance and energy saving. In China, the wastewater treatment industry started late due to the constraints of industrialization and economic development, and the wastewater treatment plants are few and unevenly distributed. Most wastewater treatment plants have disadvantages such as outdated equipment and technology, high energy consumption, low automation and intelligence, and are in urgent need of expansion, transformation and upgrading. In order to reduce energy consumption and wastewater treatment costs and improve wastewater treatment quality and wastewater treatment efficiency, China's wastewater treatment plants should develop towards intelligence, automation, precision and low energy consumption in future. The upgrading and transformation of the aeration process, which is the key energy consumption part of the wastewater treatment plant, are the top priority of future research and development. Considering the low feasibility and high cost of direct experiments in the wastewater treatment plant, modeling and simulation control are first required to provide theoretical basis for actual experiments. However, as the aeration process involves a series of tasks such as biological reactions and sludge reflux, there are many influencing factors which make the model complicated and difficult to construct. At present, the aeration process control in China usually adopts manual proportional-integral-derivative (PID) control or manual empirical control, which has low control accuracy and features uncertainty, non-linearity, time variability, time delay and large inertia. In overall, the fine control of the aeration process is a major and difficult problem in wastewater treatment. SUMMARY In order to overcome the shortcomings of the prior art, an objective of the present disclosure is to provide an adaptive control method and system for an aeration process. The present disclosure optimizes and transforms an aeration process in a biological treatment system of a wastewater treatment plant, and improves the stability and utilization of the aeration process. To achieve the above purpose, the present disclosure provides the following technical solutions. An adaptive control method for an aeration process includes: acquiring a characteristic parameter of wastewater, a characteristic parameter of activated sludge (AS) and a blowing rate of a blower at a current moment; calculating, through an adaptive control model of a dissolved oxygen (DO) concentration in an aerobic tank, an adjustment target of the DO concentration in the aerobic tank at a next moment according to the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at the current moment; and determining a blowing rate of the blower at the next moment according to the adjustment target of the DO concentration in the aerobic tank at the next moment; where the adaptive control model of the DO concentration in the aerobic tank is determined by constructing a particle swarm optimization-backpropagation (PSO-BP) neural network and an activated sludge model No. 1 (ASM1). Optionally, the acquiring a characteristic parameter of wastewater, a characteristic parameter of AS and a blowing rate of a blower at a current moment specifically includes: acquiring a real-time characteristic parameter of the wastewater by using wastewater characteristic sensors installed at a water inlet, a water outlet, an aerobic tank and a secondary settling tank, where the characteristic parameter of the wastewater includes temperature, chemical oxygen demand (COD), ammonia nitrogen, total Kjeldahl nitrogen (TKN), DO and pH; acquiring a real-time characteristic parameter of the AS by using online sludge concentration meters installed in the aerobic tank and the secondary settling tank, where the characteristic parameter of the AS is mixed liquor suspended solids (MLSS); and acquiring a real-time blowing rate of the blower by using a blowing rate sensor installed at an air outlet of an outlet pipe of the blower. Optionally, the adaptive control model of the DO concentration in the aerobic tank is determined as follows: constructing an ASM1 model; and simulating a PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASMI model as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determining the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. Optionally, the constructing an ASM Imodel specifically includes: constructing an ASMI model based on a COST624&682 benchmark simulation process flow by using Simulink or Graphical User Interface (GUI) in Matrix Laboratory (MATLAB). Optionally, the PSO-BP neural network-based adaptive control model is determined as follows: determining sample data, where the sample data includes the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at a historical moment; constructing a BP neural network; determining initial weights of a hidden layer and an output layer by using a PSO algorithm; determining a BP neural network-based proportional-integral-derivative (PID) control structure according to the BP neural network and the initial weights of the hidden layer and the output layer; and training the BP neural network-based PID control structure by the sample data to obtain a PSO-BP neural network-based adaptive control model. An adaptive control system for an aeration process includes: an information acquisition module, for acquiring a characteristic parameter of wastewater, a characteristic parameter of AS and a blowing rate of a blower at a current moment; an aerobic tank DO concentration adjustment target calculation module, for calculating, through an adaptive control model of a DO concentration in an aerobic tank, an adjustment target of the DO concentration in the aerobic tank at a next moment according to the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at the current moment; and a blowing rate determination module, for determining a blowing rate of the blower at the next moment according to the adjustment target of the DO concentration in the aerobic tank at the next moment; where the adaptive control model of the DO concentration in the aerobic tank is determined by constructing a PSO-BP neural network and an ASM1. Optionally, the information acquisition module specifically includes: a wastewater characteristic parameter acquisition unit, for acquiring a real-time characteristic parameter of the wastewater by using wastewater characteristic sensors installed at a water inlet, a water outlet, an aerobic tank and a secondary settling tank, where the characteristic parameter of the wastewater includes temperature, COD, ammonia nitrogen, TKN, DO and pH; an AS characteristic parameter acquisition unit, for acquiring a real-time characteristic parameter of the AS by using online sludge concentration meters installed in the aerobic tank and the secondary settling tank, where the characteristic parameter of the AS is MLSS; and a blowing rate acquisition unit, for acquiring a real-time blowing rate of the blower by using a blowing rate sensor installed at an air outlet of an outlet pipe of the blower. Optionally, the adaptive control system for an aeration process further includes an aerobic tank DO concentration adaptive control model construction module, where the aerobic tank DO concentration adaptive control model construction module includes: an ASM1 model construction unit, for constructing an ASM1 model; and an optimization unit, for simulating a PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASMI model as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determining the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. Optionally, the ASM Imodel construction unit specifically includes: an ASMI model construction subunit, for constructing an ASMI model based on a COST624&682 benchmark simulation process flow by using Simulink in MATLAB or GUI in MATLAB. Optionally, the optimization unit specifically includes: a sample data determination subunit, for determining sample data, where the sample data includes the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at a historical moment; a BP neural network construction subunit, for constructing a BP neural network; an initial weight calculation subunit, for determining initial weights of a hidden layer and an output layer by using a PSO algorithm; a PID control structure determination subunit, for determining a BP neural network-based PID control structure according to the BP neural network and the initial weights of the hidden layer and the output layer; a PSO-BP neural network-based adaptive control model determination subunit, for training the BP neural network-based PID control structure by the sample data to obtain a PSO-BP neural network-based adaptive control model; and an optimization subunit, for simulating the PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASMI model as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determining the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. According to the specific embodiments of the present disclosure, the present disclosure provides the following technical effects: The present disclosure optimizes the design of a controller for an aeration process of a wastewater treatment plant, and constructs an adaptive control model of a dissolved oxygen (DO) concentration in an aerobic tank based on a particle swarm optimization-backpropagation (PSO-BP) neural network. By adding the intelligent control of the DO concentration of the aerobic tank, the present disclosure breaks the limitation of traditional control based on manual experience, and improves the stability and utilization of the aeration process. BRIEF DESCRIPTION OF DRAWINGS To describe the technical solutions in embodiments of the present disclosure or in the prior art more clearly, the accompanying drawings needed in the embodiments are introduced below briefly. Apparently, the accompanying drawings in the following description show merely some embodiments of the present disclosure, and other drawings can be derived from these accompanying drawings by those of ordinary skill in the art without creative efforts. FIG. 1 is a flowchart of an adaptive control method for an aeration process according to the present disclosure. FIG. 2 is a structural diagram of an adaptive control system for an aeration process according to the present disclosure. FIG. 3 shows a COST624&682 benchmark simulation process flow according to the present disclosure. FIG. 4 shows encapsulation of each process component in a biological reaction tank according to the present disclosure. FIG. 5 shows a diagram of encapsulated modules for a sub-process reaction rate of a process component, a particle moving velocity in a secondary settling tank and an activated sludge (AS) concentration in each voxel layer in a model according to the present disclosure. FIG. 6 shows overall encapsulation of an activated sludge model No. 1 (ASM1) using an anoxic/oxic (A/O) process according to the present disclosure. FIG. 7 is a structural diagram of a backpropagation (BP) neural network according to the present disclosure. FIG. 8 shows a control structure of a particle swarm optimization-backpropagation (PSO-BP) neural network-based adaptive algorithm according to the present disclosure.
FIG. 9 shows an aeration process under PSO-BP neural network-based adaptive control according to the present disclosure. DETAILED DESCRIPTION The technical solutions of the embodiments of the present disclosure are clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are merely a part rather than all of the embodiments of the present disclosure. All other embodiments derived from the embodiments of the present disclosure by a person of ordinary skill in the art without creative efforts should fall within the protection scope of the present disclosure. An objective of the present disclosure is to provide an adaptive control method and system for an aeration process. The present disclosure optimizes and transforms an aeration process in a biological treatment system of a wastewater treatment plant, and improves the stability and utilization of the aeration process. To make the above objectives, features and advantages of the present disclosure clearer and more comprehensible, the present disclosure is described in further detail below with reference to the accompanying drawings and specific implementations. Embodiment 1 This embodiment provides an adaptive control method for an aeration process. As shown in FIG. 1, this method includes: Step 101: Acquire a characteristic parameter of wastewater, a characteristic parameter of activated sludge (AS) and a blowing rate of a blower at a current moment. Specifically: Acquire a real-time characteristic parameter of the wastewater by using wastewater characteristic sensors installed at a water inlet, a water outlet, an aerobic tank and a secondary settling tank, where the characteristic parameter of the wastewater includes temperature, chemical oxygen demand (COD), ammonia nitrogen, total Kjeldahl nitrogen (TKN), dissolved oxygen (DO) and pH. Acquire a real-time characteristic parameter of the AS by using online sludge concentration meters installed in the aerobic tank and the secondary settling tank, where the characteristic parameter of the AS is mixed liquor suspended solids (MLSS). Acquire a real-time blowing rate of the blower by using a blowing rate sensor installed at an air outlet of an outlet pipe of the blower. Step 102: Calculate, through an adaptive control model of a DO concentration in an aerobic tank, an adjustment target of the DO concentration in the aerobic tank at a next moment according to the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at the current moment. The adaptive control model of the DO concentration in the aerobic tank is determined by constructing a particle swarm optimization-backpropagation (PSO-BP) neural network and an activated sludge model No. 1 (ASM1). Step 103: Determine a blowing rate of the blower at the next moment according to the adjustment target of the DO concentration in the aerobic tank at the next moment. The adaptive control model of the DO concentration in the aerobic tank is determined as follows: Construct an ASMI model, specifically, construct an ASMI model based on a COST624&682 benchmark simulation process flow by using Simulink or Graphical User Interface (GUI) in Matrix Laboratory (MATLAB). Simulate a PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASMI model as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determine the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. The PSO-BP neural network-based adaptive control model is determined as follows: Determine sample data, where the sample data includes the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at a historical moment. Construct a BP neural network. Determine initial weights of a hidden layer and an output layer by using a PSO algorithm. Determine a BP neural network-based proportional-integral-derivative (PID) control structure according to the BP neural network and the initial weights of the hidden layer and the output layer. Train the BP neural network-based PID control structure by the sample data to obtain a PSO-BP neural network-based adaptive control model. Embodiment 2 This embodiment provides an adaptive control system for an aeration process. As shown in FIG. 2, this system includes: an information acquisition module, an aerobic tank DO concentration adjustment target calculation module and a blowing rate determination module. The information acquisition module 201 is used for acquiring a characteristic parameter of wastewater, a characteristic parameter of AS and a blowing rate of a blower at a current moment, and specifically includes: a wastewater characteristic parameter acquisition unit, for acquiring a real-time characteristic parameter of the wastewater by using wastewater characteristic sensors installed at a water inlet, a water outlet, an aerobic tank and a secondary settling tank, where the characteristic parameter of the wastewater includes temperature, COD, ammonia nitrogen, TKN, DO and pH; an AS characteristic parameter acquisition unit, for acquiring a real-time characteristic parameter of the AS by using online sludge concentration meters installed in the aerobic tank and the secondary settling tank, where the characteristic parameter of the AS is MLSS; and a blowing rate acquisition unit, for acquiring a real-time blowing rate of the blower by using a blowing rate sensor installed at an air outlet of an outlet pipe of the blower. The aerobic tank DO concentration adjustment target calculation module 202 is used for calculating, through an adaptive control model of a DO concentration in an aerobic tank, an adjustment target of the DO concentration in the aerobic tank at a next moment according to the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at the current moment, where the adaptive control model of the DO concentration in the aerobic tank is constructed based on a PSO-BP neural network and an ASMI module. The blowing rate determination module 203 is used for determining a blowing rate of the blower at the next moment according to the adjustment target of the DO concentration in the aerobic tank at the next moment. The adaptive control system for an aeration process in this embodiment further includes an aerobic tank DO concentration adaptive control model construction module. The aerobic tank DO concentration adaptive control model construction module includes: an ASM1 model construction unit, for constructing an ASM1 model; and an optimization unit, for simulating a PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASMI model as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determining the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. The ASM Imodel construction unit specifically includes: an ASMI model construction subunit, for constructing an ASMI model based on a COST624&682 benchmark simulation process flow by using Simulink in MATLAB or GUI in MATLAB. The optimization unit specifically includes: a sample data determination subunit, for determining sample data, where the sample data includes the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower at a historical moment; a BP neural network construction subunit, for constructing a BP neural network; an initial weight calculation subunit, for determining initial weights of a hidden layer and an output layer by using a PSO algorithm; a PID control structure determination subunit, for determining a BP neural network-based PID control structure according to the BP neural network and the initial weights of the hidden layer and the output layer; and a PSO-BP neural network-based adaptive control model determination subunit, for training the BP neural network-based PID control structure by the sample data to obtain a PSO-BP neural network-based adaptive control model; and an optimization subunit, for simulating the PSO-BP neural network-based adaptive control model by a Monte Carlo method by taking the DO concentration in the aerobic tank during the aeration process as a control object and the ASM Imodel as a platform to obtain an optimized PSO-BP neural network-based adaptive control model, and determining the optimized PSO-BP neural network-based adaptive control model as the adaptive control model of the DO concentration in the aerobic tank. Embodiment 3 This embodiment provides a PSO-BP neural network-based adaptive control method for an aeration process. This method includes the following steps: Step Sl: Monitor a characteristic parameter of wastewater and a characteristic parameter of AS in a water inlet, a water outlet, an aerobic tank and a secondary settling tank in real time through online sensors, and monitor a blowing rate of a blower on line through a blowing rate sensor installed at an air outlet of an air outlet pipe of the blower. Step S2: Acquire the characteristic parameter of the wastewater, the characteristic parameter of the AS and the blowing rate of the blower during the aeration process through a controller installed in a blower room, and take these variables and parameters as a current system state. Step S3: Calculate, by the controller, an optimal adjustment target of a DO concentration in an aerobic tank under the current system state through an adaptive control model of the DO concentration in the aerobic tank. Step S4: Adjust, by the controller, a blowing rate of the blower according to the adjustment target of the DO concentration in the aerobic tank so as to achieve adaptive intelligent control. In the PSO-BP neural network-based adaptive control method for an aeration process in this embodiment, the characteristic parameter of the wastewater includes temperature, COD, ammonia nitrogen, TKN, DO and pH, and the characteristic parameter of the AS is MLSS. In the PSO-BP neural network-based adaptive control method for an aeration process in this embodiment, the online sensors include wastewater characteristic sensors, AS characteristic sensors and blowing rate sensors. The wastewater characteristic sensors are installed in the water inlet, the water outlet, the aerobic tank and the secondary settling tank, including a flow meter, an online thermometer, an online level gauge, an online pH meter, an online oxidation-reduction potential (ORP) tester, an online DO meter, an online COD meter, an online ammonia nitrogen meter and an online TKN meter. The AS characteristic sensors are installed in the aerobic tank and the secondary settling tank, and are online sludge concentration meters. The blowing rate sensors are respectively installed at the air outlet of the air outlet pipe of the blower. In the PSO-BP neural network-based adaptive control method for an aeration process in this embodiment, a construction method of the adaptive control model of the DO concentration in the aerobic tank is specifically described as follows: This embodiment uses an ASM1 model proposed by the International Association on Water Quality (IAWQ). The ASMI model divides substances in the wastewater into 13 process components based on biological characteristics such as solubility and composition, and divides a system reaction process into eight sub-processes based on microbial metabolism and electron receptors. The rate of each sub-process is described by using a dual Monod model, which is a centralized parameter model with no spatial change in the model parameters. The ASMI model divides an entire biological process in a biological reaction tank into eight sub-processes (including three microbial growth processes, two microbial decay processes, one ammonification process and two hydrolysis processes). The material in the biological reaction tank is divided into 13 process components (including eight carbon and nitrogen substrates, three microbial substances, one additional electron acceptor and one process alkalinity change). Several process components participate in each sub-process, and each process component participates in several sub-processes. In addition, five stoichiometric parameters are defined for a measurement relationship between the process components in each sub-process, and 14 reaction kinetic parameters are defined for the process components participating in each sub-process reaction, which include half-saturation coefficient, microbial decay coefficient, maximum specific growth rate of microorganisms, correction factors for microbial growth and hydrolysis, and ammonification and hydrolysis rates. This embodiment sequentially establishes a sub-process reaction rate equation and a total reaction rate equation of single process components based on the ASMI model, and unifies the units of each process component to generate a final model. The wastewater treatment system is a large non-linear system that is affected by large disturbances such as influent, flow rate, pollutant load and influent wastewater composition. To optimize the control strategy of the wastewater treatment system by establishing a model, many difficulties need to be overcome, such as a large range of time constants (ranging from a few minutes to a few days), the complexity of biological reactions and the lack of standard evaluation criteria. Therefore, a qualified process benchmark is required to establish and test the ASM1 model. The ASM1 model of the present disclosure is constructed and verified based on a COST624&682 benchmark simulation process flow. The COST624&682 benchmarks are committed to reducing research costs and improving wastewater treatment technology by deepening the understanding of microorganisms in the AS system and strengthening the control and optimization of the biological treatment process. The COST624&682 benchmark simulation process flow has the following requirements: a) The model is composed of five continuous biochemical tanks (two anoxic tanks and three aerobic tanks, using an anoxic/oxic (A/O) process) and a secondary settling tank. b) A biochemical tank 1 and a biochemical tank 2 are anoxic tanks, where no air is input. A biochemical tank 3, a biochemical tank 4 and a biochemical tank 5 are aerobic tanks, in which a saturated concentration of DO is 8 g/m3 . The default values of a volumetric oxygen transfer coefficient (KLa) are 10/hr, 10/hr and 3.5/hr respectively. ASMI is used as a biochemical reaction process model. c) There is no biochemical reaction in the secondary settling tank, and the secondary settling tank is divided into 10 layers. A sedimentation process model is described by using a double exponential sedimentation rate equation proposed by Takacs. d) A mixed liquor circulates from the biochemical tank 5 to the biochemical tank 1, and AS circulates from the secondary settling tank to the biochemical tank 1. e) The standards give physical structure parameters of the biochemical tanks and the secondary settling tank, the ASMI model's kinetic parameters, stoichiometric parameters and parameter values of the secondary settling tank model at 15°C, steady state influent and effluent values of the ASM Imodel and steady state values and effluent values of each layer of the secondary settling tank, dynamic state influent and effluent values of the ASM1 model and dynamic state values and effluent values of each layer of the secondary settling tank. FIG. 3 shows a COST624&682 benchmark simulation process flow. As shown in FIG. 3, Qin represents an influent flow rate; Qr represents an amount of refluxed sludge; Qrin represents an amount in internal circulation; Qf represents a flow rate of wastewater into the secondary settling tank; Qe represents an amount of supernatant discharged from the secondary settling tank; Qw represents an amount of AS discharged from the secondary settling tank; Zin represents a concentration of an influent component; Zr represents a concentration of a component in refluxed AS; Zrin represents a concentration of a circulating component; Zf represents a concentration of a component in the wastewater flowing into the secondary settling tank; Ze represents a concentration of a component in the discharged supernatant; Zw represents a concentration of a component in the discharged AS. In addition, this embodiment further uses two other more intuitive and concise modeling representations in MATLAB, as follows: This embodiment uses Simulink in MATLAB to express the modeling more intuitively and concisely. The encapsulation of each mechanism of the ASMI model constructed by Simulink is shown in FIGS. 4 to 6. The biological reaction tank model is composed of encapsulated modules of each process component. The concentration changes of the process components follow the principle of material conservation, and the same process component participates in multiple reactions. The structure of the model constructed by Simulink is clear and intuitive, and the model operation effect can be obtained through the output value changes of an oscilloscope and operation process. This embodiment further uses GUI in MATLAB to express the modeling more intuitively and concisely. According to the designed functional modules, main functions of an interface of the simulation system include input sample selection, sampling stride setting, kinetic parameter setting, stoichiometric parameter setting and the drawing of DO concentration change curves of the three aerobic tanks. The interface is designed to include edit boxes, radio buttons, buttons and coordinate axis boxes, static text boxes and other controls. The specific design is as follows: a) Control and sampling parameter settings. Three edit boxes are used to acquire system sampling stride, control sampling stride, sampling days and other parameters. b) Input sample selection. Five radio buttons are used to select an input sample. c) Drawing of DO concentration change curves of three aerobic tanks. Two buttons are used to select program operation and start drawing a curve or clear the drawn curve in the coordinate axis, and three coordinate axis boxes are used to display the curve. d) Initial performance parameter setting. 19 static texts and 19 edit boxes are used to prompt and input the set kinetic parameters and stoichiometric parameters respectively. The interface of the ASMI model designed by GUI has functions such as parameter setting, input selection and output curve drawing. This embodiment conducts a simulation experiment on the control effect of the PSO-BP neural network-based adaptive control model by taking the DO concentration of the aerobic tank in the aeration process as a control object and the ASM Imodel as a platform. PSO is an evolutionary computing technology that treats every solution to an optimization problem as a particle. All particles are searched in a D-dimensional space to look for the best position. A fitness function is used to determine a fitness value so as to evaluate the current position. In addition, the particle continuously adjusts its velocity according to its own flowing experience and the flying experience of its companions, that is, to adjust the direction and distance of the flying. In addition, the particle can remember its currently determined optimal position during the search process. As shown in FIG. 7, this embodiment adopts a three-layer BP neural network, which includes four input nodes, five hidden nodes and three output nodes. In the figure, j represents an input layer, i represents a hidden layer, and 1 represents an output layer. yr (k), y(k), e(k), represents an input into the BP neural network, and kp ki k d represents and output from the BP neural network. A control structure of the PSO-BP neural network-based adaptive control model is shown in FIG. 8. The BP neural network-based PID control method has a problem that the initial weight is sensitive, that is, the selection of the initial weights of the hidden layer and the output layer have a great influence on the final operating result of the system. The selection of a suitable initial weight often requires a large number of repeated experiments. If the initial weight selection and test are carried out manually, the operation is too tedious and boring. In this embodiment, the PSO algorithm allows the system to perform multiple preliminary selections of initial weights by itself, and select the initial weights with the best comprehensive effect after recording the results. A specific learning process includes: a) Calculate the total number of connection weights in the BP neural network as a dimensional value of particles in the PSO algorithm. b) Randomly generate a particle population, set the number of evolutions, and perform iterative learning according to the PSO algorithm. c) Take a mean squared error (MSE) function of the entire system output as the fitness function to evaluate the generated individuals in order to survive the fittest, and calculate, through the powerful iterative search ability of the PSO algorithm, the weight of the BP neural network that minimizes the MSE of the system output. In this embodiment, the control method uses the PSO-BP neural network to train the adaptive control model of the DO concentration in the aeration process. The training parameters include learning rate r/=0.003 and inertia coefficient y=0.003. The initial weights are random values, and weights with the best operation stability are selected by using the PSO algorithm. Immunity and robustness are important indicators to measure the performance of a control system. In order to facilitate observation and analysis of experimental results, this embodiment lets the ASMI model in a steady input state and adds a Gaussian white noise with a signal-to-noise ratio (SNR) of 20 to feedback signals of an aeration tank to simulate the
PSO-BP neural network-based adaptive control model of the DO concentration as shown in FIG. 9. This embodiment uses a Monte Carlo method to test the robustness of the aeration process of the AS system in the PSO-BP neural network-based adaptive control model of the DO concentration. In order to facilitate observation and analysis of experimental results, the system is let in a steady input state. A parameter perturbation interval is selected for a yield coefficient YA of autotrophic bacteria, a yield coefficient YH of heterotrophic bacteria, a ratio fP of inert particles in microorganisms, a ratio of nitrogencontent XB in microbial cells and a ratio ixP of nitrogen content in the microbial product in the experimental model. These parameters are evenly distributed in the given interval. 50 sets of parameter values are taken for random combination, and then a simulation experiment is conducted to obtain the state of the adaptive control model when the parameters are randomly combined. Meanwhile, the upper limits of these parameters are taken to simulate to derive the adaptive control model when the parameters are severe. Each embodiment in the specification of the present disclosure is described in a progressive manner. Each embodiment focuses on the difference from other embodiments, and the same and similar parts between the embodiments may refer to each other. For the system disclosed in the embodiments, since the system corresponds to the method disclosed in the embodiments, the description is relatively simple, and reference can be made to the method description. Several embodiments are used for illustration of the principles and implementation methods of the present disclosure. The description of the embodiments is used to help illustrate the method and its core principles of the present disclosure. In addition, those skilled in the art can make various modifications in terms of specific embodiments and scope of application in accordance with the teachings of the present disclosure. In conclusion, the content of the present specification should not be construed as a limitation to the present disclosure.