Microflora prediction-based two-stage activated sludge system optimization method and system

Abstract

The invention discloses a two-stage activated sludge system optimization method and system based on microbial community prediction, and the method comprises the following steps: constructing a microbial community succession prediction model of a first stage, and obtaining the data of microbial communities in activated sludge; constructing a pollutant removal efficiency prediction model of the activated sludge system in a second stage, and predicting an effluent index of the activated sludge by taking the output index and the regulation and control index of the model in the first stage as input; and constructing a two-stage active sludge system intelligent optimization model, integrating the first-stage prediction model and the second-stage prediction model by adopting a non-dominated sorting genetic algorithm, constructing a complete two-stage microbial community optimal control model, adjusting regulation and control indexes to directionally regulate and control succession of the microbial community, and realizing the maximum removal rate of pollutants of the active sludge system. By constructing the microbial community succession prediction model and the pollutant removal efficiency prediction model, accurate regulation and optimization of the activated sludge system can be realized.

Description

Technical Field

[0001] The invention belongs to the technical field of water treatment and relates to a two-stage activated sludge system optimization method and system based on microbial community prediction. Background Art

[0002] As a widely used sewage treatment method, activated sludge technology is based on the use of flocculent sludge formed by microorganisms under aerobic conditions to effectively remove organic matter, nitrogen, phosphorus and other pollutants in sewage. Due to its high efficiency, economy and environmental friendliness, it is widely used in urban sewage and industrial wastewater treatment around the world.

[0003] Although activated sludge technology has many advantages, its stability and efficiency are facing challenges. This is mainly because the microbial community structure in activated sludge technology is extremely complex, and the interaction between different microbial species and quantities is intricate, making it difficult to effectively regulate through traditional empirical methods. Therefore, in actual operation, it is often difficult to maintain the stability and efficiency of treatment efficiency, resulting in large fluctuations in treatment effects and difficulty in meeting increasingly stringent environmental protection requirements.

[0004] To solve this problem, researchers have been exploring how to more accurately and effectively regulate the microbial community structure in activated sludge technology. However, due to the complexity and dynamics of the microbial community structure and the lack of effective quantitative regulation methods, solving this problem has become particularly difficult. Summary of the invention

[0005] The purpose of the present invention is to solve the problem of lack of effective quantitative control methods for the complexity and dynamics of microbial community structure in the prior art, and to provide a two-stage activated sludge system optimization method and system based on microbial community prediction.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The present invention proposes a two-stage activated sludge system optimization method based on microbial community prediction, comprising the following steps:

[0008] The first phase of the microbial community succession prediction model was constructed to obtain data on the microbial community in the activated sludge, and the microbial community was decomposed into diversity and abundance by combining the environmental parameters and influent data of the reactor. The time factor was incorporated to track the dynamic changes of the microbial community over time, and the state of the microbial community at a specific future moment was predicted through a time series enhancement strategy.

[0009] Construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output indicators and control indicators of the first-stage model as input to predict the effluent indicators of the activated sludge and evaluate the pollutant treatment performance of the activated sludge system;

[0010] The non-dominated sorting genetic algorithm is used in combination with intelligent algorithms to integrate the first and second stage prediction models, and an intelligent optimization model of the two-stage activated sludge system is constructed. The removal rates of chemical oxygen demand, total nitrogen and total phosphorus are taken as objective functions, and the control indicators are used as the solution space. By adjusting the control indicators, the succession of the microbial community is directionally regulated to achieve the maximum removal rate of pollutants in the activated sludge system.

[0011] Preferably, the construction of the first stage microbial community succession prediction model specifically comprises the following steps:

[0012] Data preparation and preprocessing: collect and organize time series data containing the abundance of the top 15 microbial species, clean the data to remove outliers and missing values, and perform standardization;

[0013] Multi-task neural network structure design, design a multi-task neural network, where each task corresponds to the abundance prediction of a microbial species, the network shares hidden layers to capture the potential correlation between the abundances of different species, and each task has an independent output layer for predicting the abundance of the corresponding species;

[0014] Feature selection and input: features related to microbial species abundance were selected as input, including time, water inflow index, regulation index, and the abundance of the top 15 species in the microbial community, and the features were input into the input layer of the multi-task neural network;

[0015] Model training and optimization: Use the prepared data set to train the multi-task neural network, use the mean square error as the loss function to measure the difference between the predicted result and the true value, use the optimization algorithm Adam to update the network weights to minimize the loss function, and evaluate the performance of the model through the 5-fold cross-validation method to perform hyperparameter tuning;

[0016] Model application and feedback: Apply the trained multi-task neural network to the actual operation of the activated sludge system, make real-time predictions of microbial species abundance, adjust control indicators based on the prediction results to optimize the succession process of the microbial community, and collect feedback data from actual applications to improve and optimize the model.

[0017] Preferably, during the data preparation and preprocessing, historical data are collected and sorted, including the composition and abundance of microbial communities at different time points and the corresponding water inlet indicators and control parameters; the data are cleaned, outliers and missing values ​​are removed, and standardized to ensure the consistency and comparability of the data;

[0018] The Min-Max method is used for data normalization, as shown in formula (1):

[0019]

[0020] Among them, x min is the minimum value of the variable, x max is the maximum value of the variable, x new is the normalized variable.

[0021] Preferably, the data of microbial communities in activated sludge are obtained by:

[0022] A light-sequential batch reactor of the same laboratory scale was used, the reactor was placed in a light-shielded position, and LED lights were arranged around the reactor to provide constant intensity of light, the light intensity was set to 5000±200 lux, and the light time was set to 12 hours of light / 12 hours of darkness;

[0023] The test water is introduced from the top of the reactor by a water pump. The liquid level is controlled by a liquid level relay. The drain port is located in the middle of the reactor and is controlled by a solenoid valve. The water volume is changed by 50% in each cycle.

[0024] A sand core aeration head is set at the bottom of the reactor for aeration, and the aeration rate is 2L / min;

[0025] The reactor was operated at room temperature with a cycle of 6 hours, including water inlet, aeration, sedimentation, drainage and idleness, of which the water inlet and drainage stages were 5 minutes each, the anaerobic idleness after water inlet was 30 minutes, the sedimentation time was 5 minutes, and the aeration time was 315 minutes;

[0026] Synthetic wastewater was used as the operating water source of the reactor, wherein the carbon source, nitrogen source and phosphorus source were sodium acetate, ammonium sulfate and potassium dihydrogen phosphate respectively;

[0027] The inoculated sludge is the bacterial and algal granular sludge cultured in the previous stage;

[0028] Reactors R1-R3 are designed to control the organic load rate index, R4-R6 to control the carbon-nitrogen ratio index, and R7-R9 to control the organic nitrogen content index;

[0029] Influent, effluent and biomass samples were collected, and influent indicators were measured, including chemical oxygen demand, ammonia nitrogen, total nitrogen and total phosphorus. The community data of fungal microorganisms and eukaryotic microorganisms were measured by performing 16SrRNA high-throughput sequencing and 18SrRNA high-throughput sequencing on biomass samples, respectively.

[0030] Preferably, the pollutant removal efficiency prediction model of the activated sludge system in the second stage is constructed as follows:

[0031] Data preparation, including collecting pollutant removal efficiency data of activated sludge systems, cleaning the data, removing outliers and missing values, and standardizing the data;

[0032] Feature selection, selecting features related to pollutant removal efficiency from the collected data, the features including microbial community indicators and regulation indicators;

[0033] Model training: Use the selected ensemble tree algorithm and the processed data set to train the model, adjust the model parameters, including the number and depth of trees, and optimize the model performance through cross-validation.

[0034] Preferably, the data preparation specifically includes:

[0035] Integrate the output data of the first-stage model and the actual control parameter data to form an initial data set;

[0036] Clean the data and remove outliers and missing values ​​to ensure the integrity and accuracy of the data; standardize or normalize the data to improve the efficiency and accuracy of model training;

[0037] The output data include the effluent indicators of the activated sludge system, specifically the concentrations of COD-OUT, TN-OUT, and TP-OUT pollutants; the input data include the number of fungal microbial populations and their percentage in the total microbial population, the abundance of the top 15 microbial species, and the percentage of algae microbial abundance among the top 15 species in the total abundance, and the control indicators include specifically the organic load rate, nitrogen load, and carbon-nitrogen ratio.

[0038] Preferably, the construction of the two-stage activated sludge system intelligent optimization model is specifically as follows:

[0039] Problem definition and model initialization: The goal is to solve the combination of control index parameters that can maximize the removal rate of chemical oxygen demand, total nitrogen and total phosphorus under the current microbial community state, and initialize the parameters of the non-dominated sorting genetic algorithm, including population size, genetic generation, crossover probability and mutation probability, and randomly generate the initial population representing the candidate solutions of the control index parameters;

[0040] Fitness evaluation: Substitute each individual in the initial population into the pollutant removal efficiency prediction model of the activated sludge system. The individual is a combination of control index parameters, calculate the corresponding chemical oxygen demand, total nitrogen and total phosphorus removal rates, and calculate the fitness value of each individual based on the removal rate;

[0041] Non-dominated sorting and selection: non-dominated sorting is performed on individuals in the population, and individuals are divided into different non-dominated levels according to fitness values. Parents are selected based on non-dominated levels and crowding information to generate the next generation of population;

[0042] Genetic operation: performing crossover and mutation operations on selected parent individuals to generate new offspring individuals. The crossover operation includes randomly selecting two parent individuals and exchanging some of their genes to generate two new offspring individuals. The mutation operation includes randomly selecting a gene in the offspring individuals and changing it to increase population diversity.

[0043] Population update and iteration: merge the newly generated offspring individuals with the parent individuals to form a new population, perform non-dominated sorting and crowding calculation on the new population, select individuals with higher fitness and uniform distribution as the next generation population, and repeat the non-dominated sorting and selection, genetic operation steps until the preset genetic generation is reached or the stopping condition is met;

[0044] Optimal solution selection and output: select the individuals with the highest fitness and uniform distribution from the final population as the optimal solution, and output the control index parameter combination corresponding to the optimal solution as the optimal control scheme to meet the chemical oxygen demand, total nitrogen and total phosphorus removal rate under the current microbial community state;

[0045] Verification and application: Apply the obtained optimal control index parameter combination to the actual operation of the activated sludge treatment system to verify its effect under actual conditions, and adjust and optimize the model according to the verification results to improve the prediction accuracy and actual application effect.

[0046] Preferably, the optimal solution is measured by a balanced value, and the difference between the objective function values ​​is minimized in the Pareto optimal solution set that meets the conditions, specifically:

[0047]

[0048] Wherein: Eq is the equilibrium value of the three objective functions in the Pareto optimal solution; COD is the removal rate of COD in the Pareto optimal solution; TP is the removal rate of TP in the Pareto optimal solution when COD is the Pareto optimal solution; TN is the removal rate of TN in the Pareto optimal solution when COD is the Pareto optimal solution; k is the equilibrium value coefficient.

[0049] The present invention proposes a two-stage activated sludge system optimization system based on microbial community prediction, comprising:

[0050] The first data processing module is used to construct the first-stage microbial community succession prediction model, obtain the data of the microbial community in the activated sludge, combine the environmental parameters of the reactor and the influent data, decompose the microbial community into diversity and abundance, incorporate the time factor to track the dynamic changes of the microbial community over time, and predict the state of the microbial community at a specific future moment through the time series enhancement strategy;

[0051] The second data processing module is used to construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output index and control index of the first-stage model as input, predicting the effluent index of the activated sludge, and evaluating the pollutant treatment performance of the activated sludge system;

[0052] The model optimization module is used to integrate the first-stage and second-stage prediction models using a non-dominated sorting genetic algorithm in combination with an intelligent algorithm, to construct an intelligent optimization model for a two-stage activated sludge system, with the removal rates of chemical oxygen demand, total nitrogen and total phosphorus as the objective function and the control index as the solution space. By adjusting the control index, the succession of the microbial community is directionally controlled to achieve the maximum removal rate of pollutants in the activated sludge system.

[0053] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a two-stage activated sludge system optimization method based on microbial community prediction are implemented.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] A two-stage activated sludge system optimization method based on microbial community prediction in the present invention can realize accurate regulation and optimization of the activated sludge (AS) system by constructing a microbial community succession prediction model and a pollutant removal efficiency prediction model. This helps to flexibly adjust the control index parameters according to the microbial community state and pollutant removal requirements in actual operation, thereby improving the treatment efficiency and stability of the system. Improve the pollutant removal rate, use the non-dominated sorting genetic algorithm (NSGA) to solve the optimal control index parameters, and the present invention can find a control scheme that meets the maximum removal rate of chemical oxygen demand (COD), total nitrogen (TN) and total phosphorus (TP) under the current microbial community state. This not only helps to improve the removal efficiency of pollutants, but also reduces processing costs and improves resource utilization. Intelligence and automation, the present invention combines machine learning algorithms and intelligent optimization technology to realize intelligent and automated control of the AS system. This reduces manual intervention and dependence, and improves the operating efficiency and reliability of the system. At the same time, the intelligent control method also helps to promptly discover and respond to abnormal situations in the operation of the system to ensure the stable operation of the system. Scalability and adaptability, the prediction model and control method of the present invention have good scalability and adaptability. As the operating conditions of the AS system change and the structure of the microbial community evolves, the prediction performance and control effect can be optimized by updating the data set and retraining the model. This enables the present invention to continuously adapt to different operating environments and processing requirements, and maintain its effectiveness and competitiveness in practical applications. Energy conservation, emission reduction and sustainable development, by optimizing the operating parameters and control strategies of the AS system, the present invention helps to reduce energy consumption and waste emissions, and achieve energy conservation, emission reduction and sustainable development goals. This is in line with the current global trend of environmental protection and sustainable development, and helps to promote innovation and progress in water treatment technology. Compared with the prior art, the present invention has shown significant beneficial effects in terms of precise control and optimization of the AS system, improvement of pollutant removal rate, intelligent and automated control, scalability and adaptability, energy conservation, emission reduction and sustainable development, etc., which makes the present invention have broad application prospects and important practical value in the field of water treatment.

[0056] The present invention proposes a two-stage activated sludge system optimization system based on microbial community prediction, which divides the system into a first data processing module, a second data processing module and a model optimization module, and adjusts the control index to directionally control the succession of the microbial community, thereby achieving the maximum removal rate of pollutants in the activated sludge system. The modular concept is adopted to make each module independent of each other, which is convenient for unified management of each module. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 This is a flow chart of the two-stage activated sludge system optimization method based on microbial community prediction of the present invention.

[0059] Figure 2 This is a structural diagram of the construction of the two-stage activated sludge system optimization model of the present invention.

[0060] Figure 3 This is a structural diagram of the optimal multi-task neural network for the first-stage microbial community succession prediction model of the present invention.

[0061] Figure 4 This is a diagram of the optimization system of the two-stage activated sludge system based on microbial community prediction of the present invention.

[0062] Figure 5 The figure is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0064] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0066] Example 1

[0067] The present invention proposes a two-stage activated sludge system optimization method based on microbial community prediction, such as Figure 1 As shown in the figure, the flow chart of the optimal control model of the two-stage activated sludge system is constructed, which specifically includes the following steps:

[0068] S1, construct the first stage of microbial community succession prediction model to obtain the data of microbial community in activated sludge, combine the environmental parameters and influent data of the reactor, decompose the microbial community into diversity and abundance, incorporate the time factor to track the dynamic changes of the microbial community over time, and predict the state of the microbial community at a specific future moment through the time series enhancement strategy.

[0069] The first phase of the microbial community succession prediction model was constructed, which specifically included the following steps:

[0070] S1.1, data preparation and preprocessing, collected and organized time series data containing the abundance of the top 15 microbial species, cleaned the data to remove outliers and missing values, and standardized them.

[0071] During data preparation and preprocessing, historical data are collected and organized, including the composition and abundance of microbial communities at different time points, as well as the corresponding water inlet indicators and control parameters; data are cleaned, outliers and missing values ​​are removed, and standardized to ensure data consistency and comparability;

[0072] The Min-Max method is used for data normalization, as shown in formula (1):

[0073]

[0074] Among them, x min is the minimum value of the variable, x max is the maximum value of the variable, x new is the normalized variable.

[0075] S1.2, Multi-task neural network structure design, design a multi-task neural network, where each task corresponds to the abundance prediction of a microbial species, the network shares hidden layers to capture the potential correlation between the abundances of different species, and each task has an independent output layer for predicting the abundance of the corresponding species.

[0076] S1.3, feature selection and input, select features related to microbial species abundance as input, including time, water inflow indicators, regulation indicators, and the abundance of the top 15 species in the microbial community, and input the features into the input layer of the multi-task neural network.

[0077] S1.4, model training and optimization, use the prepared data set to train the multi-task neural network, use the mean square error as the loss function to measure the difference between the predicted result and the true value, use the optimization algorithm Adam to update the network weights to minimize the loss function, and evaluate the performance of the model through the 5-fold cross-validation method to perform hyperparameter tuning.

[0078] S1.5, model application and feedback, applies the trained multi-task neural network to the actual operation of the activated sludge system, performs real-time prediction of microbial species abundance, adjusts control indicators based on the prediction results to optimize the succession process of the microbial community, and collects feedback data from actual applications to improve and optimize the model.

[0079] When obtaining data on the microbial community in activated sludge, specifically:

[0080] The same laboratory-scale photo-sequential batch reactor was used, each of which had a height of 1100 mm, an inner diameter of 70 mm, a height-to-diameter ratio of 15.7, and an effective volume of 3 L. The reactor was placed in a light-shielded position, and LED lights were arranged around it to provide constant intensity illumination, the illumination intensity was set to 5000 ± 200 lux, and the illumination time was set to 12 hours of light / 12 hours of darkness;

[0081] The test water is introduced from the top of the reactor by a water pump. The liquid level is controlled by a liquid level relay. The drain port is located in the middle of the reactor and is controlled by a solenoid valve. The water volume is changed by 50% in each cycle.

[0082] A sand core aeration head is set at the bottom of the reactor for aeration, and the aeration rate is 2L / min;

[0083] The reactor was operated at room temperature with a cycle of 6 hours, including water inlet, aeration, sedimentation, drainage and idleness, of which the water inlet and drainage stages were 5 minutes each, the anaerobic idleness after water inlet was 30 minutes, the sedimentation time was 5 minutes, and the aeration time was 315 minutes;

[0084] Synthetic wastewater was used as the operating water source of the reactor, wherein the carbon source, nitrogen source and phosphorus source were sodium acetate, ammonium sulfate and potassium dihydrogen phosphate respectively;

[0085] The inoculated sludge is the bacterial and algal granular sludge cultured in the previous stage;

[0086] Reactors R1-R3 are designed to control the organic load rate index, R4-R6 to control the carbon-nitrogen ratio index, and R7-R9 to control the organic nitrogen content index;

[0087] Influent, effluent and biomass samples were collected, and influent indicators were measured, including chemical oxygen demand, ammonia nitrogen, total nitrogen and total phosphorus. The community data of fungal microorganisms and eukaryotic microorganisms were measured by performing 16SrRNA high-throughput sequencing and 18SrRNA high-throughput sequencing on biomass samples, respectively.

[0088] The high-throughput sequencing method includes: using the Ezup column soil DNA extraction kit to extract the total genomic DNA of the sludge sample, testing the concentration and purity of the DNA, and storing it at -20°C after meeting the standards, and then sequencing it on the MiSeq platform. The primers for bacterial sequencing are 338F and 806R, and the primers for algae sequencing are 3NDF and V4-euk-R2R.

[0089] S2, construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, use the output indicators and control indicators of the first-stage model as input, predict the effluent indicators of the activated sludge, and evaluate the pollutant treatment performance of the activated sludge system.

[0090] Construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, specifically:

[0091] S2.1, data preparation, includes collecting pollutant removal efficiency data of activated sludge system, cleaning the data, removing outliers and missing values, and standardizing the data.

[0092] The data preparation is as follows:

[0093] Integrate the output data of the first-stage model and the actual control parameter data to form an initial data set;

[0094] Clean the data and remove outliers and missing values ​​to ensure the integrity and accuracy of the data; standardize or normalize the data to improve the efficiency and accuracy of model training;

[0095] The output data include the effluent indicators of the bacteria-algae granular sludge (ABGS) system, specifically the concentrations of COD-OUT, TN-OUT, and TP-OUT pollutants; the input data include the number of fungal microbial populations and their percentage in the total microbial population, the abundance of the top 15 microbial species, and the percentage of algae microbial abundance among the top 15 species in the total abundance, and the control indicators are specifically the organic load rate, nitrogen load, and carbon-nitrogen ratio.

[0096] It should be noted that in order to reflect the application scope of the present invention as much as possible, this example applies an activated sludge improvement technology with a more complex microbial community than that of the activated sludge system: the "ABGS system".

[0097] S2.2, feature selection, selects features related to pollutant removal efficiency from the collected data, and the features include microbial community indicators and regulation indicators.

[0098] S2.3, model training, train the model using the selected ensemble tree algorithm and the processed data set, adjust the model parameters, including the number and depth of trees, and optimize the performance of the model through cross-validation.

[0099] S3, combined with intelligent algorithms, uses a non-dominated sorting genetic algorithm to integrate the first-stage and second-stage prediction models, and constructs a complete two-stage activated sludge system intelligent optimization model. The removal rates of chemical oxygen demand, total nitrogen and total phosphorus are taken as objective functions, and the control indicators are used as the solution space. By adjusting the control indicators, the succession of microbial communities is directionally controlled to achieve the maximum removal rate of pollutants in the activated sludge system.

[0100] Construct an intelligent optimization model for a two-stage activated sludge system, specifically:

[0101] S3.1, problem definition and model initialization, the goal is to solve the control index parameter combination that can maximize the chemical oxygen demand, total nitrogen and total phosphorus removal rate under the current microbial community state, and initialize the parameters of the non-dominated sorting genetic algorithm, including population size, genetic generation, crossover probability and mutation probability, and randomly generate an initial population representing the candidate solutions of the control index parameters.

[0102] S3.2, fitness evaluation, substitute each individual in the initial population into the pollutant removal efficiency prediction model of the activated sludge system, the individual is the combination of control index parameters, calculate the corresponding chemical oxygen demand, total nitrogen and total phosphorus removal rates, and calculate the fitness value of each individual based on the removal rate.

[0103] S3.3, non-dominated sorting and selection, performs non-dominated sorting on individuals in the population, divides individuals into different non-dominated levels according to their fitness values, and selects parents based on the non-dominated level and crowding information to generate the next generation of the population.

[0104] S3.4, genetic operation, performs crossover and mutation operations on the selected parent individuals to generate new offspring individuals, where the crossover operation includes randomly selecting two parent individuals and exchanging some of their genes to generate two new offspring individuals, and the mutation operation includes randomly selecting a gene in the offspring individuals and changing it to increase population diversity.

[0105] S3.5, population update and iteration, merge the newly generated offspring individuals with the parent individuals to form a new population, perform non-dominated sorting and crowding calculation on the new population, select individuals with higher fitness and uniform distribution as the next generation population, and repeat the non-dominated sorting and selection, genetic operation steps until the preset genetic generation is reached or the stopping condition is met.

[0106] S3.6, optimal solution selection and output, select the individuals with the highest fitness and uniform distribution from the final population as the optimal solution, and output the control index parameter combination corresponding to the optimal solution as the optimal control scheme to meet the chemical oxygen demand, total nitrogen and total phosphorus removal rate under the current microbial community state.

[0107] The optimal solution is measured by the equilibrium value. In the Pareto optimal solution set that meets the conditions, the difference between the objective function values ​​is minimized. Specifically:

[0108]

[0109] Wherein: Eq is the equilibrium value of the three objective functions in the Pareto optimal solution; COD is the removal rate of COD in the Pareto optimal solution; TP is the removal rate of TP in the Pareto optimal solution when COD is the Pareto optimal solution; TN is the removal rate of TN in the Pareto optimal solution when COD is the Pareto optimal solution; k is the equilibrium value coefficient.

[0110] S3.7, verification and application, apply the obtained optimal control index parameter combination to the actual operation of the activated sludge treatment system, verify its effect under actual conditions, and adjust and optimize the model according to the verification results to improve the prediction accuracy and actual application effect.

[0111] Example

[0112] The model structure framework diagram of an embodiment of the present invention is as follows Figure 2 As shown, the specific steps are as follows:

[0113] Step 1: Construction of the first phase of microbial community succession prediction model:

[0114] The data of microbial communities in activated sludge are obtained, combined with the environmental parameters and influent data of the reactor. The model exhaustively decomposes the microbial community into two main aspects: diversity and abundance, and cleverly incorporates the time factor to accurately track the dynamic changes of the microbial community over time. By applying the time series enhancement strategy, the model can accurately predict the state of the microbial community at a specific future moment based on the initial conditions, time span, influent indicators and control parameters.

[0115] The microbial community succession prediction model is built using a support vector machine and a multi-task neural network algorithm.

[0116] The diversity characteristics of the microbial community include: the number of fungal microbial populations and their proportion; the abundance characteristics of the microbial community include: the abundance of the top 15 microbial species and the proportion of algae microbial abundance and other key microbial indicators.

[0117] The output indicators of the first-stage model of microbial community succession prediction model include: microbial community diversity characteristics and abundance characteristics, and the input indicators include: data of past time nodes of output indicators, the time interval from past time nodes to the present, the influent indicators of the activated sludge system and the indicators that can be regulated.

[0118] In order to demonstrate the effect of the present invention, the activated sludge process provided is selected as a case of bacterial-algal granular sludge (ABGS) with a more complex microbial community. First, the microbial community data and other environmental indicator data related to the model established by the present invention should be collected.

[0119] The data of microbial communities in activated sludge are obtained, combined with the environmental parameters and influent data of the reactor, including:

[0120] For data acquisition, this example used nine identical laboratory-scale photo-sequencing batch reactors (PSBR).

[0121] Specifically, the photo-sequencing batch reactor (PSBR) has a height of 1100 mm, an inner diameter of 70 mm, a height-to-diameter ratio (H / D) of 15.7, and an effective volume of 3 L. During operation, the reactor is placed in a light-shielding position and LED lights are set around it to provide constant intensity of light. The test water enters the reactor from the top through a water pump, the liquid level is controlled by a liquid level relay, the drain port is located in the middle of the reactor, and is controlled by a solenoid valve. The water volume is changed by 50% and the water volume is 1.5 L in each cycle. A sand core aeration head is set at the bottom of the reactor for aeration, and the aeration volume is controlled at 2 L / min by a glass rotor flowmeter. The PSBR reactor is operated at room temperature, with an operating cycle of 6 h. The operating time of each stage is adjusted by a time controller, including five stages: water inlet, aeration, sedimentation, drainage, and idle. Among them, water inlet and drainage are both 5 min, anaerobic idle for 30 min after water inlet, sedimentation time is 5 min, and aeration time is 315 min. An LED lamp was used as a light source and placed at the back of the reactor. The light intensity was set to 5000±200 lux near the inner wall of the reactor, and the illumination time was set to 12h light / 12h dark. The reactor was operated using synthetic wastewater.

[0122] The carbon source, nitrogen source and phosphorus source are sodium acetate, ammonium sulfate and potassium dihydrogen phosphate respectively. The inoculated sludge is the ABGS cultured in the previous stage.

[0123] In order to study the effects of different operating conditions on ABGS cultivation, referring to the operating indicators often involved in the experiment, this study designed reactors R1-R3 to regulate the organic loading rate (OLR) indicator, R4-R6 to regulate the carbon-nitrogen ratio (C / N) indicator, and R7-R9 to regulate the organic nitrogen content (ON) indicator, and called them the regulation indicators. For the initial inoculation of ABGS, the reactors of the three regulation indicators in this study were taken from different batches of sludge cultivated by this research group in the early stage. Influent, effluent and biomass samples were collected every day to measure the influent indicators (chemical oxygen demand (COD-IN), ammonia nitrogen (NH3-IN), total nitrogen (TN-IN), total phosphorus (TP-IN)) and the bacterial and algal microbial community of ABGS (16SrRNA high throughput, 18SrRNA high throughput) at different times of the reactor. 16SrRNA high throughput measures fungal microorganisms, 18SrRNA high throughput measures eukaryotic microorganisms, and 18SrRNA high throughput results also include algal microorganisms. The remaining indicators were measured according to standard methods. To facilitate the construction of the model,

[0124] Specifically, the high-throughput sequencing method is: the total genomic DNA of the sludge sample is extracted using the Ezup column soil DNA extraction kit of Sangon Biotech (Shanghai) Co., Ltd., and then the concentration and purity of the DNA are detected using a Nandrop2000 micro-spectrophotometer. The DNA samples that meet the standards are stored at -20°C. Sequencing was performed on the MiSeq platform (Illumina, USA) by Paisono Biotechnology Co., Ltd. (Shanghai, China). The upstream primer for bacterial sequencing is 338F (5'-ACTCCTACGGGAGGCAGCAG-3'), and the downstream primer is 806R (5'-GGACTACH VGGGTWTCTAAT-3'). The upstream primer for algae sequencing is 3NDF (5'-GGCAAGTCTGGTGCCAG-3'), and the downstream primer is V4-euk-R2R (5'-ACGGTATCTRATCRTCTTCG-3').

[0125] The pollutant removal efficiency index of ABGS is set as the removal of carbon (C), nitrogen (N) and phosphorus (P), which are expressed by the removal rates of COD, TN and TP respectively (COD-OUT, TN-OUT, TP-OUT (%)). For the convenience of explanation, they are called effluent indexes. For the convenience of model visualization, some indicators are redefined and described, and their detailed explanations are shown in Table 1. The range of all collected data is shown in Table 2.

[0126] Table 1 Description of the features present in the dataset used for modeling

[0127]

[0128]

[0129] Table 2 Index range of reactor

[0130]

[0131] Specifically, in step 1, according to the initial conditions, time span, water inlet index and control parameters, the present invention provides a prediction method based on a multi-task neural network algorithm to accurately predict the state of the microbial community at a specific future moment, specifically including:

[0132] S11, Data preparation and preprocessing

[0133] Specifically, historical data are collected and organized, including the composition and abundance of microbial communities at different time points, as well as the corresponding water inlet indicators and control parameters.

[0134] Specifically, the data were cleaned, outliers and missing values ​​were removed, and standardized to ensure data consistency and comparability.

[0135] It should be noted that since the dimensions of each influencing factor are different, each dimension is normalized. In this paper, the Min-Max method is used for data normalization, as shown in formula (1).

[0136] S12, Multi-task neural network model construction

[0137] Specifically, a multi-task neural network structure is designed, in which each task corresponds to the abundance prediction of one or more microbial species, or to the prediction of dynamic changes of different functional microbial communities.

[0138] Specifically, the network shares the underlying hidden layers to capture the potential correlations between different tasks, while each task has an independent output layer to predict the future state of the corresponding microbial community.

[0139] Specifically, the output indicators include: the number of fungal microbial populations (16S) and their percentage in the total microbial population (PCT of 16S), the abundance of the top 15 microbial species (Ab), and the percentage of algae microbial abundance among the top 15 species in the total abundance (PCT of 18S). The input variables include the status and time of the previous time node of the output indicator, influent indicators (chemical oxygen demand (COD-IN), ammonia nitrogen (NH3-IN), total nitrogen (TN-IN), total phosphorus (TP-IN)) and control indicators (OLR, ON, C / N).

[0140] S13, model training and optimization

[0141] Specifically, the multi-task neural network model is trained using the prepared dataset, and the network weights are updated through the back-propagation algorithm and optimizer (Adam).

[0142] Specifically, 5-fold cross validation and L2 regularization were used to prevent overfitting, and Bayesian optimization was applied to adjust hyperparameters such as network structure (such as the number of layers and neurons) to optimize model performance.

[0143] The final optimized hyperparameters are: fully connected layers: 2, first layer size: 15, second layer size: 17, activation function: none, regularization strength (Lambda): 5.96*10^-8, normalized data: yes. The specific model structure of the multi-task neural network can be found in Figure 3 .

[0144] S15, Prediction and Verification

[0145] Specifically, the trained model is applied to new data sets or ABGS systems in actual operation to make future predictions of the state of the microbial community. The prediction accuracy and reliability of the model are verified by comparison with actual observational data.

[0146] S16, Results Analysis and Application

[0147] Specifically, the prediction results were analyzed to identify the dynamic trends of key microbial species or functional communities. The model performance was evaluated by the coefficient of determination (R2) and root mean square error (RMSE) indicators.

[0148] The performance of the microbial community succession prediction model in the first stage was: RMSR = 102.004, R2 = 0.951.

[0149] In summary, the present invention realizes the accurate prediction of the succession of microbial communities in the ABGS system through a multi-task neural network algorithm combined with initial conditions, time span, water inlet indicators and control parameters, providing strong support for the intelligent control and optimization of the system.

[0150] Step 2, the pollutant removal efficiency prediction model of the second stage activated sludge system, including:

[0151] The model takes the output indicators and control indicators of the first-stage model as input and directly predicts the effluent indicators of activated sludge, thereby comprehensively evaluating the pollutant treatment performance of the activated sludge system.

[0152] The pollutant removal efficiency prediction model of the activated sludge system is constructed using an integrated tree algorithm.

[0153] This patent uses the removal rates of chemical oxygen demand (COD), total nitrogen (TN) and total phosphorus (TP) to characterize the pollutant treatment performance of activated sludge and use them as output indicators of the second-stage model.

[0154] When constructing the effluent index prediction model of the activated sludge system (specifically the algae granular sludge ABGS system), the output index of the first stage model (such as microbial community diversity, abundance change) and the control index (OLR, C / N, ON) are used as input features, and the ensemble tree (TE) algorithm is used for training to accurately predict the effluent index of the ABGS system, such as the concentration of effluent COD, TN, and TP pollutants. This step specifically includes:

[0155] S21, Data Integration and Preprocessing

[0156] Specifically, the output data of the first-stage model and the actual control parameter data are integrated to form a training data set.

[0157] Specifically, the data is cleaned to remove outliers and missing values ​​to ensure the integrity and accuracy of the data. The data is standardized or normalized to improve the efficiency and accuracy of model training.

[0158] Specifically, the output indicators include: ABGS system effluent indicators (COD-OUT, TN-OUT, TP-OUT) and pollutant concentrations. Input variables include: fungal microbial population (16S) and its percentage of the total microbial population (PCT of16S), the abundance of the top 15 microbial species (Ab), and the percentage of algae microbial abundance in the top 15 species to the total abundance (PCT of 18S) and control indicators (OLR, ON, C / N).

[0159] S22, Ensemble Tree Algorithm Selection and Configuration

[0160] According to the complexity of the problem and the characteristics of the data, choose an appropriate ensemble tree algorithm, such as random forest, gradient boosted tree (GBDT), extreme gradient boosting (XGBoost), etc.

[0161] Specifically, configure algorithm parameters, such as the number of trees, tree depth, learning rate, etc., to optimize the prediction performance of the model.

[0162] For example, the final optimized hyper parameters of the concentration prediction model of pollutants in the effluent indicators (COD-OUT, TN-OUT, TP-OUT) of the ABGS system are shown in Table 3.

[0163] Table 3 Hyperparameters of the optimal machine learning algorithms after Bayesian optimization for seven models

[0164]

[0165] S23, model training

[0166] Specifically, the preprocessed data set is used to train the ensemble tree model, and the model parameters are iteratively updated to make the model gradually approach the actual changes in water discharge indicators.

[0167] Specifically, during the training process, 5-fold cross validation and regularization are used to prevent model overfitting and improve the generalization ability of the model.

[0168] S24, Model Validation and Optimization

[0169] Specifically, an independent validation dataset was used to evaluate the prediction performance of the model, and indicators such as the coefficient of determination (R2) and root mean square error (RMSE) were calculated.

[0170] Specifically, the COD-OUT prediction model is: RMSE = 0.314, R 2 =0.980, TN-OUT prediction model is: RMSE = 1.288, R 2 =0.944, while the TP-OUT prediction model is: RMSE = 0.200, R 2 =0.983.

[0171] S25, Model Application and Feedback

[0172] Specifically, the trained ensemble tree model is applied to the actual operation of the ABGS system to perform real-time prediction of water output indicators.

[0173] In summary, the present invention uses the output indicators and control indicators of the first-stage model as input, and uses the integrated tree algorithm to train a model for predicting the effluent indicators of activated sludge, thereby achieving accurate prediction of the effluent quality of the ABGS system, providing strong support for the intelligent operation and performance optimization of the system.

[0174] Step 3, construction of intelligent optimization model of two-stage activated sludge system:

[0175] On the basis of step 1 and step 2, combined with the advantages of intelligent algorithms in solving the optimal solution, the present invention uses a non-dominated sorting genetic algorithm (NSGA) to integrate the two-stage prediction model and construct a complete two-stage microbial community optimization model. The model takes the removal rate of COD, TN and TP as the objective function, and the control index as the solution space. By finely adjusting the control index, the succession of the microbial community is directional controlled, and the maximum removal rate of pollutants in the activated sludge system is achieved.

[0176] As an example, in order to demonstrate the ultimate function of the present invention, four ultimate pollutant removal requirements are selected, namely: balanced and efficient removal of carbon, nitrogen and phosphorus, efficient removal of carbon and nitrogen, efficient removal of nitrogen and phosphorus, and efficient removal of carbon and phosphorus.

[0177] When constructing the optimal control model of the two-stage activated sludge system (specifically the bacterial algae granular sludge ABGS system), the removal rates of COD, TN and TP are taken as the objective functions, and the control indexes (OLR, C / N, ON) are taken as the solution space. The non-dominated sorting genetic algorithm (NSGA) is used to integrate and optimize the two-stage prediction model to establish a two-stage activated sludge system optimal control model that can comprehensively consider the succession of microbial communities and the removal efficiency of pollutants. This step specifically includes:

[0178] S31, Objective function definition

[0179] Specifically, the objective function was clarified, i.e., maximizing the removal rates of COD, TN, and TP to reflect the treatment efficiency of the ABGS system.

[0180] Specifically, the pollutant removal performance prediction model of the second stage activated sludge in step 2 is used as the objective function.

[0181] S32, solving spatial definition

[0182] Specifically, determine the value range of the control index to form a solution space. Select the NSGA algorithm as the solution method because it can effectively handle multi-objective optimization problems and find a set of non-dominated solutions (i.e., Pareto optimal solutions). Configure algorithm parameters such as population size, genetic generation, crossover probability, mutation probability, etc. to optimize the algorithm's search efficiency and performance.

[0183] For example, the solution space of the present invention is the upper and lower bounds of the control indicators (OLR, C / N, ON), as shown in Table 2.

[0184] S33 Two-stage prediction model integration

[0185] Specifically, the first-stage microbial community succession prediction model and the second-stage effluent index prediction model were integrated to form a complete two-stage prediction model.

[0186] In the NSGA algorithm, the prediction models of the three pollutant removal rates of the activated sludge prediction model in the second stage are output as the basis for fitness evaluation.

[0187] Given that the ABGS system needs to meet three objectives in terms of pollutant removal efficiency, the optimal solution obtained by the NSGA algorithm is not a single solution, but a series of Pareto optimal solutions (PS) that meet the conditions. In order to evaluate the pros and cons of each solution in the PS, this study considered the strict requirement that the removal rates of the three pollutants must meet the standards in actual engineering applications, that is, in the PS, the smaller the difference between the three objective function values, the more it can reflect the characteristics of achieving the optimal solution at the same time. Therefore, this study uses the degree of difference between the three objective variable values ​​in the PS to measure the pros and cons of the solution, and defines it as the equilibrium value (Eq), see formula (2) for details.

[0188] Specifically, where: Eq is the equilibrium value of the three objective functions in the Pareto optimal solution; COD is the removal rate of COD in the Pareto optimal solution; TP is the removal rate of TP in the Pareto optimal solution when COD is the Pareto optimal solution; TN is the removal rate of TN in the Pareto optimal solution when COD is the Pareto optimal solution; k is the equilibrium value coefficient, which is set to 4 in this study for the convenience of analyzing Eq.

[0189] S34, genetic manipulation and fitness evaluation

[0190] Specifically, genetic operations such as crossover and mutation are performed on individuals in the population (i.e., combinations of control indicators) to generate new offspring individuals.

[0191] Specifically, a two-stage prediction model was used to evaluate the fitness of offspring individuals and calculate their corresponding COD, TN and TP removal rates.

[0192] S35, Nondominated Sorting and Selection

[0193] Specifically, individuals in the population are non-dominated and sorted according to their fitness values ​​to form different non-dominated hierarchies.

[0194] Specifically, in each level, individuals with higher fitness and uniform distribution are selected as candidate solutions for the next generation according to the crowding information.

[0195] S36, Iterative Optimization and Generation of Pareto Optimal Solution Sets

[0196] Specifically, genetic operations, fitness evaluation, and non-dominated sorting are repeated until a preset genetic generation number is reached or a stopping condition is met.

[0197] The hyperparameters of NSGA in this embodiment are set as follows: selection operation used: tournament selection, optimal front-end individual coefficient: 0.25, population size: 1000, maximum evolutionary generation: 1000, stop generation: 1000, fitness function value deviation: 1e-100.

[0198] Specifically, through the above parameter settings, 250 Pareto optimal solution sets (PS) that meet the requirements can be selected.

[0199] Specifically, the Pareto optimal solution set is output, that is, a set of optimal control index combinations in a given solution space, which balance the removal rates of COD, TN and TP to different degrees.

[0200] S37, Optimal Control Model Application and Feedback

[0201] Specifically, the combination of control indicators in the Pareto optimal solution set is applied to the actual operation of the ABGS system for performance verification and optimization.

[0202] The model of the present invention simulates the control strategies and final control results of four different engineering requirements, and the specific settings are as follows: the EQ project sets the goal of C, N, and P removal rates as balanced as possible, and the high-efficiency removal rate is a secondary goal. The lowest points of Eq selected from 250 PS are: COD-OUT, TN-OUT, TP-OUT, and Eq are 87.16%, 88.11%, 89.97%, and 3.50, respectively. This method can make the removal rate of each pollutant reach more than 85%. The NC project sets the goal of N and P removal rates as efficient as possible, and C removal rate is a secondary goal. The lowest points of the balanced values ​​selected from 250 PS are: COD-OUT, TN-OUT, TP-OUT, and Eq are 81.25%, 90.76%, 94.00%, and 16.23, respectively. This method can make the N and P removal rates reach more than 90%. The NN project sets the goal of C and P removal as efficient as possible, and N removal is a secondary goal. The lowest points of the equilibrium values ​​selected from 250 PS are: COD-OUT, TN-OUT, TP-OUT, Eq are 90.08%, 81.01%, 90.54%, and 13.17, respectively. This method can achieve C and P removal rates above 90%. The NP project sets the goal of C and N removal as efficient as possible, and P removal is a secondary goal. The lowest points of Eq selected from 250 PS are: COD-OUT, TN-OUT, TP-OUT, and Eq are 90.00%, 90.84%, 74.09%, and 23.12, respectively. This method can achieve C and N removal rates above 90%. The results of the simulated influent treatment strategy in this study show that the NC effluent COD content is high, about 95 mg / L; the NN effluent TN content is high, about 14 mg / L; the NP effluent TP content is high, about 2.3 mg / L. The specific control results and strategies are shown in Table 4.

[0203] Table 4 NSGA-3 regulation output results under specific conditions

[0204]

[0205] According to different industrial effluent standards, engineers can choose different microbial community regulation strategies to ensure that the pollutant removal efficiency of the ABGS system remains efficient and stable. The two-stage microbial community optimization model framework transforms the empirical operation of the ABGS system into scientific operation, providing a reliable intelligent guidance model for the operation of the ABGS system.

[0206] In summary, the present invention integrates and optimizes the two-stage prediction model by adopting the non-dominated sorting genetic algorithm (NSGA), and establishes a two-stage activated sludge system optimization model that can comprehensively consider the microbial community succession and pollutant removal efficiency, providing strong support for the intelligent operation and performance optimization of the ABGS system.

[0207] Example 2

[0208] The present invention proposes a two-stage activated sludge system optimization system based on microbial community prediction, such as Figure 4 As shown, including:

[0209] The first data processing module is used to construct the first-stage microbial community succession prediction model, obtain the data of the microbial community in the activated sludge, combine the environmental parameters of the reactor and the influent data, decompose the microbial community into diversity and abundance, incorporate the time factor to track the dynamic changes of the microbial community over time, and predict the state of the microbial community at a specific future moment through the time series enhancement strategy;

[0210] The second data processing module is used to construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output index and control index of the first-stage model as input, predicting the effluent index of the activated sludge, and evaluating the pollutant treatment performance of the activated sludge system;

[0211] The model optimization module is used to integrate the first-stage and second-stage prediction models using a non-dominated sorting genetic algorithm in combination with an intelligent algorithm, to construct an intelligent optimization model for a two-stage activated sludge system, with the removal rates of chemical oxygen demand, total nitrogen and total phosphorus as the objective function and the control index as the solution space. By adjusting the control index, the succession of the microbial community is directionally controlled to achieve the maximum removal rate of pollutants in the activated sludge system.

[0212] Example 3

[0213] See also Figure 5As shown, the present invention also provides an electronic device 100 for a two-stage activated sludge system optimization method based on microbial community prediction; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0214] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the two-stage activated sludge system optimization method based on microbial community prediction described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0215] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0216] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a two-stage activated sludge system optimization method based on microbial community prediction, and the processor 102 can execute the plurality of instructions to implement:

[0217] The first phase of the microbial community succession prediction model was constructed to obtain data on the microbial community in the activated sludge, and the microbial community was decomposed into diversity and abundance by combining the environmental parameters and influent data of the reactor. The time factor was incorporated to track the dynamic changes of the microbial community over time, and the state of the microbial community at a specific future moment was predicted through a time series enhancement strategy.

[0218] Construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output indicators and control indicators of the first-stage model as input to predict the effluent indicators of the activated sludge and evaluate the pollutant treatment performance of the activated sludge system;

[0219] The non-dominated sorting genetic algorithm is used in combination with intelligent algorithms to integrate the first and second stage prediction models, and an intelligent optimization model of the two-stage activated sludge system is constructed. The removal rates of chemical oxygen demand, total nitrogen and total phosphorus are taken as objective functions, and the control indicators are used as the solution space. By adjusting the control indicators, the succession of the microbial community is directionally regulated to achieve the maximum removal rate of pollutants in the activated sludge system.

[0220] Example 4

[0221] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0222] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0223] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0224] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0225] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0226] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A two-stage activated sludge system optimization method based on microbial community prediction, characterized in that: The following steps are involved: The first phase of the microbial community succession prediction model was constructed to obtain data on the microbial community in the activated sludge, and the microbial community was decomposed into diversity and abundance by combining the environmental parameters and influent data of the reactor. The time factor was incorporated to track the dynamic changes of the microbial community over time, and the state of the microbial community at a specific future moment was predicted through a time series enhancement strategy. Construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output indicators and control indicators of the first-stage model as input to predict the effluent indicators of the activated sludge and evaluate the pollutant treatment performance of the activated sludge system; The non-dominated sorting genetic algorithm is used in combination with intelligent algorithms to integrate the first and second stage prediction models, and an intelligent optimization model of the two-stage activated sludge system is constructed. The removal rates of chemical oxygen demand, total nitrogen and total phosphorus are taken as objective functions, and the control indicators are used as the solution space. By adjusting the control indicators, the succession of the microbial community is directionally regulated to achieve the maximum removal rate of pollutants in the activated sludge system.

2. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 1, characterized in that: The construction of the first phase of the microbial community succession prediction model specifically includes the following steps: Data preparation and preprocessing: collect and organize time series data containing the abundance of the top 15 microbial species, clean the data to remove outliers and missing values, and perform standardization; Multi-task neural network structure design, design a multi-task neural network, where each task corresponds to the abundance prediction of a microbial species, the network shares hidden layers to capture the potential correlation between the abundances of different species, and each task has an independent output layer for predicting the abundance of the corresponding species; Feature selection and input: features related to microbial species abundance were selected as input, including time, water inflow index, regulation index, and the abundance of the top 15 species in the microbial community, and the features were input into the input layer of the multi-task neural network; Model training and optimization: Use the prepared data set to train the multi-task neural network, use the mean square error as the loss function to measure the difference between the predicted result and the true value, use the optimization algorithm Adam to update the network weights to minimize the loss function, and evaluate the performance of the model through the 5-fold cross-validation method to perform hyperparameter tuning; Model application and feedback: Apply the trained multi-task neural network to the actual operation of the activated sludge system, make real-time predictions of microbial species abundance, adjust control indicators based on the prediction results to optimize the succession process of the microbial community, and collect feedback data from actual applications to improve and optimize the model.

3. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 2, characterized in that: During the data preparation and preprocessing, historical data are collected and sorted, including the composition and abundance of microbial communities at different time points, as well as the corresponding water inlet indicators and control parameters; the data are cleaned, outliers and missing values ​​are removed, and standardized to ensure the consistency and comparability of the data; The Min-Max method is used for data normalization, as shown in formula (1): Among them, x min is the minimum value of the variable, x max is the maximum value of the variable, x new is the normalized variable.

4. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 1, characterized in that: The data of microbial communities in the activated sludge are obtained as follows: A light-sequential batch reactor of the same laboratory scale was used, the reactor was placed in a light-shielded position, and LED lights were arranged around the reactor to provide constant intensity of light, the light intensity was set to 5000±200 lux, and the light time was set to 12 hours of light / 12 hours of darkness; The test water is introduced from the top of the reactor by a water pump. The liquid level is controlled by a liquid level relay. The drain port is located in the middle of the reactor and is controlled by a solenoid valve. The water volume is changed by 50% in each cycle. A sand core aeration head is set at the bottom of the reactor for aeration, and the aeration rate is 2L / min; The reactor was operated at room temperature with a cycle of 6 hours, including water inlet, aeration, sedimentation, drainage and idleness, of which the water inlet and drainage stages were 5 minutes each, the anaerobic idleness after water inlet was 30 minutes, the sedimentation time was 5 minutes, and the aeration time was 315 minutes; Synthetic wastewater was used as the operating water source of the reactor, wherein the carbon source, nitrogen source and phosphorus source were sodium acetate, ammonium sulfate and potassium dihydrogen phosphate respectively; The inoculated sludge is the bacterial and algal granular sludge cultured in the previous stage; Reactors R1-R3 are designed to control the organic load rate index, R4-R6 to control the carbon-nitrogen ratio index, and R7-R9 to control the organic nitrogen content index; Influent, effluent and biomass samples were collected, and influent indicators were measured, including chemical oxygen demand, ammonia nitrogen, total nitrogen and total phosphorus. The community data of fungal microorganisms and eukaryotic microorganisms were measured by performing 16SrRNA high-throughput sequencing and 18SrRNA high-throughput sequencing on biomass samples, respectively.

5. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 1, characterized in that: The pollutant removal efficiency prediction model of the activated sludge system in the second stage is constructed as follows: Data preparation, including collecting pollutant removal efficiency data of activated sludge systems, cleaning the data, removing outliers and missing values, and standardizing the data; Feature selection, selecting features related to pollutant removal efficiency from the collected data, the features including microbial community indicators and regulation indicators; Model training: Use the selected ensemble tree algorithm and the processed data set to train the model, adjust the model parameters, including the number and depth of trees, and optimize the model performance through cross-validation.

6. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 5, characterized in that: The data preparation is specifically as follows: Integrate the output data of the first-stage model and the actual control parameter data to form an initial data set; Clean the data and remove outliers and missing values ​​to ensure the integrity and accuracy of the data; standardize or normalize the data to improve the efficiency and accuracy of model training; The output data include the effluent indicators of the activated sludge system, specifically the concentrations of COD-OUT, TN-OUT, and TP-OUT pollutants; the input data include the number of fungal microbial populations and their percentage in the total microbial population, the abundance of the top 15 microbial species, and the percentage of algae microbial abundance among the top 15 species in the total abundance, and the control indicators include specifically the organic load rate, nitrogen load, and carbon-nitrogen ratio.

7. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 1, characterized in that: The construction of the two-stage activated sludge system intelligent optimization model is specifically as follows: Problem definition and model initialization: The goal is to solve the combination of control index parameters that can maximize the removal rate of chemical oxygen demand, total nitrogen and total phosphorus under the current microbial community state, and initialize the parameters of the non-dominated sorting genetic algorithm, including population size, genetic generation, crossover probability and mutation probability, and randomly generate the initial population representing the candidate solutions of the control index parameters; Fitness evaluation: Substitute each individual in the initial population into the pollutant removal efficiency prediction model of the activated sludge system. The individual is a combination of control index parameters, calculate the corresponding chemical oxygen demand, total nitrogen and total phosphorus removal rates, and calculate the fitness value of each individual based on the removal rate; Non-dominated sorting and selection: non-dominated sorting is performed on individuals in the population, and individuals are divided into different non-dominated levels according to fitness values. Parents are selected based on non-dominated levels and crowding information to generate the next generation of population; Genetic operation: performing crossover and mutation operations on selected parent individuals to generate new offspring individuals. The crossover operation includes randomly selecting two parent individuals and exchanging some of their genes to generate two new offspring individuals. The mutation operation includes randomly selecting a gene in the offspring individuals and changing it to increase population diversity. Population update and iteration: merge the newly generated offspring individuals with the parent individuals to form a new population, perform non-dominated sorting and crowding calculation on the new population, select individuals with higher fitness and uniform distribution as the next generation population, and repeat the non-dominated sorting and selection, genetic operation steps until the preset genetic generation is reached or the stopping condition is met; Optimal solution selection and output: select the individuals with the highest fitness and uniform distribution from the final population as the optimal solution, and output the control index parameter combination corresponding to the optimal solution as the optimal control scheme to meet the chemical oxygen demand, total nitrogen and total phosphorus removal rate under the current microbial community state; Verification and application: Apply the obtained optimal control index parameter combination to the actual operation of the activated sludge treatment system to verify its effect under actual conditions, and adjust and optimize the model according to the verification results to improve the prediction accuracy and actual application effect.

8. The two-stage activated sludge system optimization method based on microbial community prediction according to claim 7, characterized in that: The optimal solution is measured by the equilibrium value, and the difference between the objective function values ​​is minimized in the Pareto optimal solution set that meets the conditions, specifically: Wherein: Eq is the equilibrium value of the three objective functions in the Pareto optimal solution; COD is the removal rate of COD in the Pareto optimal solution; TP is the removal rate of TP in the Pareto optimal solution when COD is the Pareto optimal solution; TN is the removal rate of TN in the Pareto optimal solution when COD is the Pareto optimal solution; k is the equilibrium value coefficient.

9. A two-stage activated sludge system optimization system based on microbial community prediction, characterized in that: include: The first data processing module is used to construct the first-stage microbial community succession prediction model, obtain the data of the microbial community in the activated sludge, combine the environmental parameters of the reactor and the influent data, decompose the microbial community into diversity and abundance, incorporate the time factor to track the dynamic changes of the microbial community over time, and predict the state of the microbial community at a specific future moment through the time series enhancement strategy; The second data processing module is used to construct a pollutant removal efficiency prediction model for the second-stage activated sludge system, using the output index and control index of the first-stage model as input, predicting the effluent index of the activated sludge, and evaluating the pollutant treatment performance of the activated sludge system; The model optimization module is used to integrate the first-stage and second-stage prediction models using a non-dominated sorting genetic algorithm in combination with an intelligent algorithm, to construct an intelligent optimization model for a two-stage activated sludge system, with the removal rates of chemical oxygen demand, total nitrogen and total phosphorus as the objective function and the control index as the solution space. By adjusting the control index, the succession of the microbial community is directionally controlled to achieve the maximum removal rate of pollutants in the activated sludge system.

10. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the two-stage activated sludge system optimization method based on microbial community prediction as described in any one of claims 1 to 8 are implemented.

Images