Method for improving biochemical treatment efficiency of sewage based on intelligent biological multiplication

Abstract

This invention discloses a method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication, belonging to the field of wastewater biochemical treatment technology. The invention screens an initial population of dominant microorganisms suitable for wastewater from a pre-set environmental microbial strain library, cultivates highly active microorganisms using a stratified microbial activation multiplier, collects data to construct a biological multiplication state dataset, and uses an improved LSTM model to predict changes in microbial activity and population ratio, dynamically adjusting multiplier parameters. Then, highly active microorganisms are added to a biochemical tank, and data is collected to construct a wastewater treatment state dataset. An improved random forest model is used to predict water quality changes, dynamically adjusting the microbial addition rate and biochemical tank operating parameters. A three-dimensional linkage control system is established, executing control operations in stages, and initiating effluent discharge when water quality continuously meets standards and predictions are stable. This invention improves wastewater treatment efficiency through the synergistic control of intelligent biological multiplication and dual AI prediction models, and has broad application value.

Description

Technical Field

[0001] This invention relates to the field of wastewater biochemical treatment technology, specifically to a method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication. Background Technology

[0002] Wastewater treatment plants generally face prominent problems such as large fluctuations in influent water quality, high content of recalcitrant organic matter, rapid decline in the activity of sludge in biological systems, insufficient nitrogen and phosphorus removal efficiency, and high energy and chemical consumption. This creates an urgent need for efficient, stable, low-carbon, and intelligent operation. Under these circumstances, biomultiplication technology has become the mainstream technical route for enhancing wastewater biochemical treatment. For example, The Application of BDP (Biological Double-Efficiency Process) in the Upgrading of Sewage Treatment Plant. Technology of Water Treatment. 2023, 49(11):142-146 points out that biomultiplication technology has the advantages of good process treatment effect, stable effluent quality, and strong resistance to shock loads. However, existing biomultiplication wastewater biochemical treatment technologies still struggle to achieve in-situ real-time quantification of microbial activity and bacterial community ratio. Process control often relies on post-treatment remediation, resulting in delayed response and insufficient precision, becoming a key bottleneck restricting the improvement of biochemical treatment efficiency. For example:

[0003] Chinese patent (publication number CN211284058U) discloses an integrated intelligent biological multiplication sewage treatment device. This device integrates microbial domestication and sewage treatment through multi-stage biological multiplication tanks and sensors, simplifying the system structure. However, this device has a mechanical layered structure and cannot achieve predictive dynamic control. It still belongs to passive monitoring and post-event adjustment.

[0004] Chinese patent (publication number CN117105413A) discloses a method for treating domestic sewage from highway service areas in cold regions using a diversified process of bio-multiplication coupled with combined aeration MBR. This method solves the problems of difficult nitrogen and phosphorus removal and poor load resistance under low temperature conditions in cold regions by combining bio-multiplication with combined aeration MBR, and achieves a high removal rate in low temperature scenarios. However, this method is a fixed process combination and device integration, which is only suitable for fixed operating conditions and does not have predictive dynamic adjustment capabilities, and cannot achieve synergistic control of bio-multiplication and sewage treatment.

[0005] Chinese patent (publication number CN118235148B) discloses a wastewater biological treatment process reconstruction method and system based on machine learning. This method optimizes process operating parameters through machine learning to improve treatment stability. However, this method only optimizes the process of the biological treatment tank and does not design an independent biological multiplication optimization function, so it cannot achieve optimized regulation of bacterial community activity and has insufficient source enhancement capabilities.

[0006] In summary, existing technologies generally suffer from problems such as the inability to quantify microbial activity and community ratio online, the lack of optimization methods for biological multiplication, and the absence of intelligent control strategies for wastewater biochemical treatment.

[0007] Therefore, there is an urgent need for a wastewater biochemical treatment efficiency improvement method that can simultaneously meet the engineering requirements of efficient bacterial cultivation, accurate prediction, dynamic regulation, and stable compliance, in order to break through the technical bottlenecks of traditional process regulation that are slow, inaccurate, and unstable. Summary of the Invention

[0008] Based on the aforementioned technical problems, this application discloses a method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication, specifically including:

[0009] Intelligent biological multiplication is carried out by building a stratified microbial activation multiplier and introducing an initial population of dominant microorganisms adapted to the wastewater to be treated.

[0010] The integrated sensor built into the hierarchical microbial activation multiplier collects data from the multiplier. Based on the collected data, the microbial activity and community ratio are calculated, and a biological multiplication status dataset is constructed by combining the collected data.

[0011] Based on the biomultiplication state dataset, an improved LSTM biomultiplication state prediction model is used to predict the changing trends of microbial activity and community ratio within the multiplier. Based on these trends, the parameters of the multiplier are dynamically adjusted to obtain highly active microorganisms.

[0012] Wastewater is treated biologically by adding highly active microorganisms to the biological treatment tank and conducting regular water quality monitoring to obtain a wastewater treatment status dataset.

[0013] Based on the wastewater treatment status dataset, an improved random forest wastewater treatment status prediction model is used to predict the water quality change trend in the biological treatment tank; based on the water quality change trend, the microbial dosing rate and the operating parameters of the biological treatment tank are dynamically adjusted.

[0014] When the indicators meet the preset emission standards for multiple consecutive tests and the model predicts that the water quality will remain stable and meet the standards in the future, the effluent discharge procedure will be initiated to complete the biological treatment of the wastewater.

[0015] Preferably, the layered microbial activation multiplier is specifically divided into an upper activation chamber, a middle proliferation chamber, and a lower maturation chamber from top to bottom. The orderly flow of the microbial community is achieved between the chambers through an annular inclined flow guide component. Each chamber of the multiplier is independently equipped with integrated sensors, including a temperature sensor, a dissolved oxygen sensor, a nutrient concentration sensor, and an ultrasonic sensor. The multiplier also integrates a constant temperature heating module, a micro-nano aeration module, and a nutrient slow release module to provide a controllable environment for microbial activation and proliferation.

[0016] Preferably, the biomultiplication state dataset is specifically defined as follows: by integrating sensors, key parameters of each chamber of the multiplier are collected at a preset frequency, including at least the temperature, dissolved oxygen concentration, nutrient concentration and microbial concentration of each chamber, and the bioactivity and microbial community ratio are calculated; wherein the bioactivity is calculated based on the microbial concentration collected by the ultrasonic sensor and the dehydrogenase activity detection method is used; the biomultiplication state dataset is constructed by combining the collected data and the calculated data.

[0017] Preferably, the improved LSTM biomultiplication state prediction model specifically involves: inputting a biomultiplication state dataset, using an improved LSTM algorithm to output prediction results at a preset frequency, including predicted values ​​of microbial activity and the proportion of dominant strains in each chamber; outputting the prediction results to an intelligent control terminal, comparing them with preset target thresholds, and generating targeted dynamic control decisions for multiplier parameters based on the predicted trends.

[0018] Preferably, the improved LSTM algorithm specifically involves: weighting the hidden layer outputs at different times using an attention mechanism, assigning higher weights to parameters that significantly affect microbial activity and bacterial community proportion, as shown in the formula:

[0019]

[0020]

[0021] in, for Attention weight at any moment To score attention, , These are the attention weight matrix and the attention bias term, respectively. for Predicted output at time step The time series length of the biological doubling state dataset. For time steps, for The hidden layer output at each moment;

[0022] Using a genetic algorithm, a random combination of hyperparameters is generated as the initial population of individuals. The fitness function is the model prediction error, and individual selection is performed using a roulette wheel selection method, with the following formula:

[0023]

[0024]

[0025]

[0026] in, For model prediction error, To predict the sample size, For the first The actual value of each sample For the first The predicted value for each sample, For the fitness function, Choose the probability for the individual. Population size;

[0027] A single-point crossover method is adopted, randomly selecting a crossover point and exchanging partial gene fragments of two parent individuals to generate offspring individuals. The mutation method of the offspring individuals is to randomly select individual gene positions for flipping. When the number of iterations of the genetic algorithm reaches a preset value or the fitness function value tends to stabilize, the hyperparameter combination corresponding to the individual with the highest fitness is output. Based on this hyperparameter combination, the improved LSTM is iteratively optimized.

[0028] Preferably, the dynamic control decision-making specifically involves: based on preset high activity standard thresholds, microbial community ratio adaptation standard thresholds, and synergistic effect standard thresholds, combined with predicted values, making a judgment; when the predicted values ​​meet the target thresholds, maintaining the current multiplier operating parameters and continuously monitoring the data; when the predicted values ​​do not meet the target thresholds, activating the parameter adjustment mechanism to adjust the operating parameters of constant temperature heating, aeration intensity, and nutrient supply.

[0029] Preferably, the wastewater treatment status dataset specifically comprises: adding highly active microorganisms using a multi-point uniform addition method, with the addition points evenly distributed in the upper, middle, and lower layers of the biological treatment tank; periodically collecting samples of the influent, effluent, and mixed liquor within the biological treatment tank, and monitoring parameters including chemical oxygen demand (COD) and five-day biochemical oxygen demand (BOD5). ammonia nitrogen Total phosphorus (TP) and microbial concentration Microbial activity (MA); integrate all monitoring data and construct a wastewater treatment status dataset in a three-dimensional structure of time-monitoring parameters-reaction conditions.

[0030] Preferably, the improved random forest wastewater treatment status prediction model specifically involves: inputting a wastewater treatment status dataset, and using an improved random forest algorithm to output future water quality change trends, including COD, , The predicted value of TP; when the deviation between the predicted water quality parameter and the target value is greater than the target water quality threshold, the dynamic adjustment of water quality treatment parameters is initiated. Based on the water quality prediction trend, the corresponding parameters are adjusted to ensure that the water quality is maintained within the target range.

[0031] Preferably, the improved random forest algorithm specifically involves: sampling the wastewater treatment status dataset using a stratified sampling method, and assigning higher sampling weights to low-frequency water quality anomaly samples;

[0032] A dual-objective optimization function is constructed to optimize both water quality prediction accuracy and error convergence speed. The optimal hyperparameter combination is iteratively sought using a Bayesian optimization algorithm for four core hyperparameters: number of decision trees, tree depth, minimum number of split samples, and minimum number of leaf node samples. The formula for the dual-objective optimization function is as follows:

[0033]

[0034] in, Fitness values ​​optimized for Bayesian approaches. Weighting coefficients for water quality prediction accuracy optimized by Bayes. For the number of test samples, , The first The predicted value and the actual value of each test sample. This refers to the error convergence rate;

[0035] Based on the optimal hyperparameter combination, a forest of multiple decision trees is constructed. The weighted Gini coefficient is used as the decision tree splitting rule, assigning higher splitting weights to key features affecting water quality prediction accuracy. The formula is as follows:

[0036]

[0037] in, This is the weighted Gini coefficient; the smaller the value, the higher the node purity and the better the splitting effect. The first step is to set the degree of influence of characteristics on water quality prediction. Weighting factors for water quality characteristics For the number of water quality parameter categories, For the first Under class characteristics, the first The sample proportion of water quality parameters of the same type;

[0038] In the multi-decision-tree fusion stage, a time-series weighting factor is used to assign higher voting weights to the prediction results of the decision trees corresponding to recent wastewater treatment status data. The prediction results of all decision trees are then weighted and fused to obtain the predicted value of the target water quality parameter, as shown in the formula:

[0039]

[0040] in, These are predicted values ​​for water quality parameters. For the number of decision trees, For the first The time-series weighting factor of the time step. For the first The decision tree in the th Predicted values ​​at time steps The total length of the time series. For time steps.

[0041] Preferably, the dynamic control of the water quality treatment parameters specifically involves: establishing a three-dimensional linkage system of microbial addition, biochemical tank operation, and nutrient replenishment; classifying control priorities based on prediction deviations; and performing control operations in stages according to control priorities.

[0042] Among them, the primary regulation adjusts the rate of addition of highly active microorganisms, the proportion of addition points and the addition cycle, prioritizing the rapid improvement of water quality through the action of microorganisms;

[0043] Secondary regulation adjusts the operating parameters of the biological treatment tank to optimize the microbial living environment;

[0044] The three-level regulation linkage activates the precise nutrient replenishment module, which calculates the deficiencies in carbon, nitrogen, and phosphorus sources based on prediction deviations and provides targeted supplementation of nutrients suitable for the growth of microbial communities.

[0045] Compared with the prior art, the technical solution of this application has the following technical effects:

[0046] This invention employs a layered microbial activation multiplier, which uses a three-stage structure of upper activation, middle proliferation, and lower maturation. Combined with integrated sensors for temperature, dissolved oxygen, nutrients, and ultrasound, it enables real-time calculation and precise control of microbial activity and colony ratio, significantly improving the efficiency of microbial cultivation and the quality of addition.

[0047] This invention achieves advance prediction and dynamic control of microbial activity and community ratio through an improved LSTM biomultiplication state prediction model. Based on predictive feedforward control, it enables more precise adjustment of multiplier operating parameters and faster response, significantly improving the intelligence level of the microbial multiplication process.

[0048] This invention constructs an improved random forest water quality prediction model. Through stratified sampling and weighted Gini coefficient optimization, it solves the problems of water quality data imbalance and low accuracy of abnormal sample identification, significantly improves the prediction reliability of low-frequency and high-impact water quality fluctuations, and achieves accurate prediction of multiple indicators and short cycles of wastewater quality.

[0049] This invention utilizes a three-dimensional linkage and hierarchical control mechanism involving microbial addition, biochemical tank operation, and nutrient replenishment. By adjusting according to priority and combining precise quantitative formulas with adaptive control logic, it ensures stable effluent quality while reducing energy and chemical consumption for aeration and nutrient addition, thus achieving efficient, low-carbon, and intelligent wastewater treatment.

[0050] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0051] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0053] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0054] Figure 1 The overall flow chart is for a wastewater biochemical treatment efficiency improvement method based on intelligent biological multiplication.

[0055] Figure 2 This is an architecture diagram of a wastewater biochemical treatment efficiency improvement method based on intelligent biological multiplication.

[0056] Figure 3 This is a diagram of the architecture of the improved LSTM biological doubling state prediction model in this application;

[0057] Figure 4 This is a diagram of the architecture of the improved random forest wastewater treatment status prediction model in this application;

[0058] Figure 5 This is a process data diagram of biological multiplication based on this method in the embodiments of this application;

[0059] Figure 6 This is a data diagram of the wastewater treatment process based on this method in the embodiments of this application;

[0060] Figure 7 This is a comparison chart of the biomultiplication process data of each method in the embodiments of this application;

[0061] Figure 8 This is a comparison chart of wastewater purification process data for each method in the embodiments of this application;

[0062] Figure 9 This is a comparison chart of the overall performance of the methods in the embodiments of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0064] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0065] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0066] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0067] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0068] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0069] Example 1 mainly describes a method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication, such as... Figure 1 , Figure 2 As shown, it specifically includes:

[0070] Intelligent biological multiplication is carried out by building a stratified microbial activation multiplier and introducing an initial population of dominant microorganisms adapted to the wastewater to be treated.

[0071] The integrated sensor built into the hierarchical microbial activation multiplier collects data from the multiplier. Based on the collected data, the microbial activity and community ratio are calculated, and a biological multiplication status dataset is constructed by combining the collected data.

[0072] Based on the biomultiplication state dataset, an improved LSTM biomultiplication state prediction model is used to predict the changing trends of microbial activity and community ratio within the multiplier. Based on these trends, the parameters of the multiplier are dynamically adjusted to obtain highly active microorganisms.

[0073] Wastewater is treated biologically by adding highly active microorganisms to the biological treatment tank and conducting regular water quality monitoring to obtain a wastewater treatment status dataset.

[0074] Based on the wastewater treatment status dataset, an improved random forest wastewater treatment status prediction model is used to predict the water quality change trend in the biological treatment tank; based on the water quality change trend, the microbial dosing rate and the operating parameters of the biological treatment tank are dynamically adjusted.

[0075] When the indicators meet the preset emission standards for multiple consecutive tests and the model predicts that the water quality will remain stable and meet the standards in the future, the effluent discharge procedure will be initiated to complete the biological treatment of the wastewater.

[0076] Furthermore, the screening of the initial microbial community of the dominant microorganisms adapted to the wastewater to be treated is specifically as follows: samples of the wastewater to be treated are collected and their characteristic parameters are detected. Combined with a pre-set environmental microbial strain library, high-throughput screening technology is used to screen candidate microbial strains that can tolerate wastewater toxicity and efficiently degrade target pollutants. After synergistic detection, a composite microecological community is constructed. Then, the composite microecological community is placed in an environment simulating the wastewater to be treated for directional domestication to obtain a dominant microbial community with strong adaptability and high activity.

[0077] Furthermore, the layered microbial activation multiplier is specifically designed as follows: the multiplier is divided into an upper activation chamber, a middle proliferation chamber, and a lower maturation chamber from top to bottom, with the orderly flow of the microbial community between the chambers achieved through an annular inclined flow guide component; each chamber of the multiplier is independently equipped with integrated sensors, including a temperature sensor, a dissolved oxygen sensor, a nutrient concentration sensor, and an ultrasonic sensor; the multiplier also integrates a constant temperature heating module, a micro-nano aeration module, and a nutrient slow release module to provide a controllable environment for microbial activation and proliferation.

[0078] Furthermore, the process of using the stratified microbial activation multiplier is as follows: the pre-activated dominant microbial initial population is arranged according to... The concentration is evenly added to the upper activation chamber of the multiplier, and at the same time, a sterile activation solution suitable for bacterial growth is added to ensure a reasonable volume ratio between the bacterial population and the activation solution; the stirring component of the upper activation chamber is started to fully mix the initial bacterial population and the activation solution.

[0079] After the initial activation of the upper layer is completed, the flow guiding component is activated to allow the microbial community in the upper activation chamber to slowly flow into the middle layer proliferation chamber. The micro-nano aeration module and stirring component in the middle layer proliferation chamber are then activated to increase the aeration intensity to meet the needs of rapid microbial community proliferation.

[0080] Once the bacterial community concentration in the middle layer of the proliferation chamber reaches the preset threshold, the bacterial community is introduced into the lower layer of the maturation chamber through the flow guiding component to maintain a suitable dissolved oxygen concentration and nutrient supply in the lower layer of the maturation chamber, and to carry out bacterial community stability cultivation.

[0081] Furthermore, the biomultiplication state dataset specifically comprises: acquiring key parameters of each chamber of the multiplier at a preset frequency using integrated sensors, including at least the temperature, dissolved oxygen concentration (including carbon, nitrogen, and phosphorus source concentrations), nutrient concentration, and microbial concentration of each chamber, and calculating bioactivity and microbial community ratio; wherein bioactivity is calculated based on the microbial concentration acquired by the ultrasonic sensor using a dehydrogenase activity detection method, with the formula as follows:

[0082]

[0083] in, For microbial activity, This represents the dehydrogenase activity value. Microbial concentration, The baseline activity value is set at MA ≥ 0.8 to determine that the bacterial community activity meets the standard, and MA ≥ 0.9 to meet the high activity standard.

[0084] The proportions of the microbial community were calculated using 16S rRNA sequencing technology to analyze the composition of the microbial community within the multiplier, and to calculate the proportions of dominant strains and cooperating strains. The formula is as follows:

[0085]

[0086]

[0087] in, , These represent the number of dominant strains and synergistic strains, respectively. The percentage of dominant strains, , These are the activities of the dominant strain and the synergistic strain, respectively. This refers to the effective volume of the chamber corresponding to the multiplier. , The standard volumes for the dominant strain and the synergistic strain are respectively ( When the bacterial count is ≥0.7, the bacterial community ratio is considered to meet the standard.

[0088] By combining the collected and calculated data, a biological doubling state dataset is constructed.

[0089] Furthermore, the collection of biological activity is specifically as follows: the ultrasonic sensor emits ultrasonic waves of a specific frequency into the multiplier, and by utilizing the reflection and scattering characteristics of ultrasonic waves by microbial cells, the detected sound wave signal is converted into an electrical signal and transmitted to the intelligent control terminal. The terminal automatically converts the electrical signal into microbial data.

[0090] Furthermore, such as Figure 3 The diagram shows the architecture of the improved LSTM biomultiplication state prediction model. Specifically, the improved LSTM biomultiplication state prediction model works as follows: inputting a biomultiplication state dataset, the improved LSTM algorithm outputs prediction results at a preset frequency, including the predicted microbial activity value and the predicted proportion of dominant strains in each chamber; the prediction results are output to the intelligent control terminal, compared with the preset target threshold, and based on the predicted trend, a targeted dynamic control decision for multiplier parameters is generated.

[0091] Furthermore, the improved LSTM algorithm specifically involves: weighting the hidden layer outputs at different time points using an attention mechanism, assigning higher weights to parameters that significantly affect microbial activity and community proportion, as shown in the formula:

[0092]

[0093]

[0094] in, for Attention weight at any moment To score attention, , These are the attention weight matrix and the attention bias term, respectively. for Predicted output at time step The time series length of the biological doubling state dataset. For time steps, for The hidden layer output at each moment;

[0095] Using a genetic algorithm, a random combination of hyperparameters is generated as the initial population of individuals. The fitness function is the model prediction error, and individual selection is performed using a roulette wheel selection method, with the following formula:

[0096]

[0097]

[0098]

[0099] in, For model prediction error, To predict the sample size, For the first The actual value of each sample For the first The predicted value for each sample, For the fitness function, Choose the probability for the individual. Population size;

[0100] A single-point crossover method is adopted, randomly selecting a crossover point and exchanging partial gene fragments of two parent individuals to generate offspring individuals. The mutation method of the offspring individuals is to randomly select individual gene positions for flipping. When the number of iterations of the genetic algorithm reaches a preset value or the fitness function value tends to stabilize, the hyperparameter combination corresponding to the individual with the highest fitness is output. Based on this hyperparameter combination, the improved LSTM is iteratively optimized.

[0101] Furthermore, the input layer of the improved LSTM receives a preprocessed dataset of biological doubling states, with a 9-dimensional input dimension, corresponding to temperature, dissolved oxygen concentration, and nutrient concentrations (including carbon, nitrogen, and phosphorus source concentrations). Nine characteristic parameters were identified, including microbial concentration, microbial activity, percentage of dominant strains, and percentage of synergistic strains.

[0102] The hidden layer consists of three substructures: a forget gate, an input gate, and an output gate. The number of nodes in the hidden layer ranges from 32 to 128 and is determined by optimization using a genetic algorithm. The activation function of the substructures is sigmoid, and the output of the hidden layer uses the tanh activation function.

[0103] The attention layer contains an attention weight calculation unit and a weighted fusion unit, and has a built-in attention weight matrix (dimension [1, number of hidden layer nodes]) and a bias term (dimension 1) for the hidden layer output. Perform weighted processing;

[0104] The output layer receives the weighted output results from the attention layer. The output dimension is 3-dimensional, corresponding to the predicted values ​​of microbial activity, the predicted value of the proportion of dominant strains, and the predicted value of the proportion of synergistic strains.

[0105] Furthermore, dynamic control decision-making specifically involves: based on preset high-activity standard thresholds, microbial community ratio adaptation standard thresholds, and synergistic effect standard thresholds, combined with predicted values, a judgment is made. When the predicted values ​​meet the target thresholds, the current multiplier operating parameters are maintained, and data is continuously monitored; when the predicted values ​​do not meet the target thresholds, a parameter adjustment mechanism is activated to adjust the operating parameters of constant temperature heating, aeration intensity, and nutrient supply, using the following formula:

[0106]

[0107] in, For the first The adjustment ratio of various parameters, For the first The adjustment coefficient of the parameter, These are the model's predicted values. For the target value, This is the baseline prediction deviation value.

[0108] Furthermore, the first Adjustment coefficient of various parameters Based on the degree of influence of this parameter on microbial activity (MA) and bacterial community ratio (PR), one possible reference value is: aeration intensity ( Dissolved oxygen has the greatest impact on microbial activity. =0.4; constant temperature heating ( ) Affects the rate of microbial metabolism, taking =0.3; Nutrient supply affects microbial proliferation ( )Pick =0.3.

[0109] Furthermore, the wastewater treatment status dataset specifically includes: the addition of highly active microorganisms using a multi-point uniform dosing method, with dosing points evenly distributed across the upper, middle, and lower layers of the biological treatment tank; regular collection of influent, effluent, and mixed liquor samples from the biological treatment tank; and monitoring parameters including chemical oxygen demand (COD) and five-day biochemical oxygen demand (BOD5). ammonia nitrogen Total phosphorus (TP) and microbial concentration Microbial activity (MA); integrate all monitoring data and construct a wastewater treatment status dataset in a three-dimensional structure of time-monitoring parameters-reaction conditions.

[0110] Furthermore, the wastewater treatment status dataset is Z-score standardized. Based on past wastewater treatment status datasets, it is divided into training and testing sets in a 7:3 ratio. The training set is used for model training, and the testing set is used to verify the model accuracy.

[0111] Furthermore, such as Figure 4 The diagram shown illustrates the architecture of the improved random forest wastewater treatment status prediction model. Specifically, the improved random forest model takes a wastewater treatment status dataset as input, applies an improved random forest algorithm, and outputs future water quality change trends, including COD, etc. , The predicted value of TP; when the deviation between the predicted water quality parameter and the target value is greater than the target water quality threshold, the dynamic adjustment of water quality treatment parameters is initiated. Based on the water quality prediction trend, the corresponding parameters are adjusted to ensure that the water quality is maintained within the target range.

[0112] Furthermore, the improved random forest algorithm specifically involves sampling the wastewater treatment status dataset using a stratified sampling method, assigning higher sampling weights to low-frequency water quality anomaly samples (to avoid model prediction bias caused by uneven sample distribution).

[0113] A dual-objective optimization function is constructed to optimize both water quality prediction accuracy and error convergence speed. The optimal hyperparameter combination is iteratively sought using a Bayesian optimization algorithm for four core hyperparameters: number of decision trees, tree depth, minimum number of split samples, and minimum number of leaf node samples. The formula for the dual-objective optimization function is as follows:

[0114]

[0115] in, Fitness values ​​optimized for Bayesian approaches. Weighting coefficients for water quality prediction accuracy optimized by Bayes. For the number of test samples, , The first The predicted value and the actual value of each test sample. This refers to the error convergence rate;

[0116] Based on the optimal hyperparameter combination, a forest of multiple decision trees is constructed. The weighted Gini coefficient is used as the decision tree splitting rule, assigning higher splitting weights to key features affecting water quality prediction accuracy. The formula is as follows:

[0117]

[0118] in, This is the weighted Gini coefficient; the smaller the value, the higher the node purity and the better the splitting effect. The first step is to set the degree of influence of characteristics on water quality prediction. Weighting factors for water quality characteristics For the number of water quality parameter categories, For the first Under class characteristics, the first The sample proportion of water quality parameters of the same type;

[0119] In the multi-decision-tree fusion stage, a time-series weighting factor is used to assign higher voting weights to the prediction results of the decision trees corresponding to recent wastewater treatment status data. The prediction results of all decision trees are then weighted and fused to obtain the final predicted water quality parameters, as shown in the formula:

[0120]

[0121] in, These are predicted values ​​for water quality parameters. For the number of decision trees, For the first The time-series weighting factor of the time step. For the first The decision tree in the th Predicted values ​​at time steps.

[0122] Furthermore, the dynamic control of water quality treatment parameters involves: establishing a three-dimensional linkage system of microbial dosing, biochemical tank operation, and nutrient replenishment; prioritizing control based on prediction deviations; and performing control operations in stages according to the control priority.

[0123] Among them, the primary control adjusts the injection rate, injection point ratio, and injection cycle of highly active microorganisms to prioritize rapid water quality improvement through microbial action. The formula is as follows:

[0124]

[0125] in, To adjust the rate of microbial inoculation after regulation, is the microbial inoculation rate adjustment coefficient (taken as [0.3, 0.6]). To predict bias, Standard prediction bias;

[0126] Secondary regulation synchronously adjusts the operating parameters of the biological treatment tank to optimize the microbial living environment. The formula is:

[0127]

[0128] in, The adjustment ratio for the operating parameters of the biological treatment tank. This is the adjustment coefficient for the operating parameters of the biochemical tank (taken as [0.2, 0.4]).

[0129] The three-level regulation linkage activates the precise nutrient replenishment module, which calculates the deficiencies in carbon, nitrogen, and phosphorus sources based on prediction deviations and provides targeted supplementation of nutrients suitable for the growth of microbial communities.

[0130] This embodiment details a method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication. It involves screening an initial population of dominant microorganisms suitable for the wastewater from a pre-set environmental microbial strain library, cultivating highly active microorganisms using a stratified microbial activation multiplier, collecting data to construct a biological multiplication status dataset, and using an improved LSTM model to predict changes in microbial activity and population ratio, while dynamically adjusting multiplier parameters. The highly active microorganisms are then added to a biochemical tank, and data is collected to construct a wastewater treatment status dataset. An improved random forest model is used to predict water quality changes, dynamically adjusting the microbial addition rate and biochemical tank operating parameters. A three-dimensional linkage control system is established, with tiered control operations executed. Wastewater discharge is initiated when water quality continuously meets standards and is predicted to be stable.

[0131] Example 2, based on Example 1, describes in detail the process of improving the efficiency of wastewater biochemical treatment using this method in the application scenario of an integrated wastewater treatment plant in an industrial park. The specific process is as follows:

[0132] The industrial park's integrated wastewater treatment plant is designed to treat 1000 m³ / d of wastewater, with influent water quality ranging from 380 to 450 mg / L for COD, 45 to 58 mg / L for ammonia nitrogen, and 5.2 to 6.8 mg / L for total phosphorus (TP).

[0133] Before implementing this method, the layered microbial activation multiplier was debugged to ensure the flow guiding components, constant temperature heating module, micro-nano aeration module, and stirring components of the upper activation chamber, middle proliferation chamber, and lower maturation chamber were properly configured. Based on the past three years of operating data of the wastewater treatment plant, the improved LSTM biological multiplication state prediction model and the improved random forest wastewater treatment state prediction model were pre-trained. The biological treatment tank of the wastewater treatment plant was cleaned, and the wastewater to be treated was injected to the preset level. The initial operating parameters of the biological treatment tank were adjusted to (temperature 28℃, aeration intensity 500m³ / h). Water quality monitoring points were set up to ensure that the sampling points for influent, effluent, and mixed liquor in the tank were uniform, and the monitoring frequency was set to once every 15 minutes.

[0134] To implement this method, wastewater samples were collected from the industrial park, and their pH, COD, and ammonia nitrogen parameters were measured. Combined with a pre-set environmental microbial strain library, high-throughput screening technology was used to screen out candidate microbial strains that could tolerate wastewater toxicity and efficiently degrade target pollutants. After synergistic detection, a composite microecological community was constructed. This composite microecological community was placed in an environment simulating the wastewater to be treated for 7 days for directional acclimatization, resulting in an initial community of dominant microorganisms with strong adaptability and high activity.

[0135] The initial population of dominant microorganisms after domestication was divided into... The concentration was evenly added to the upper activation chamber, along with a sterile activation solution suitable for bacterial growth. The stirring assembly was activated to thoroughly mix the bacteria with the activation solution, and activation was performed for 24 hours. The annular inclined flow guide assembly was then activated to slowly guide the microbial community from the upper activation chamber into the middle proliferation chamber. The micro-nano aeration module and stirring assembly were then activated to increase the aeration intensity to [value missing]. Proliferate for 48 hours; wait until the bacterial concentration in the middle layer of the proliferation cavity reaches... The bacteria are introduced into the lower maturation chamber through the flow guide component, maintaining a dissolved oxygen concentration of 2-3 mg / L and a suitable supply of nutrients. After 24 hours of cultivation, highly active microorganisms are obtained.

[0136] During the biomultiplication process, temperature, dissolved oxygen concentration, nutrient concentration (carbon source, nitrogen source, phosphorus source), and microbial concentration were collected from each chamber of the multiplier. After Z-score normalization preprocessing, the microbial activity (MA) was calculated using the dehydrogenase activity detection method, and the dominant strain ratio (PR) was calculated using 16S rRNA sequencing technology to construct a biomultiplication state dataset. This dataset was input into an improved LSTM biomultiplication state prediction model. This model, optimized using attention mechanism and genetic algorithm, outputs predicted values ​​of microbial activity and dominant strain ratio for each chamber within the next hour. These values ​​were compared with preset thresholds (MA≥0.9, PR≥0.7), and the constant temperature heating (maintained at 28±1℃), aeration intensity, and nutrient supply were dynamically adjusted to ensure high microbial activity. The biomultiplication process data were statistically analyzed, and the predicted and measured data during the biomultiplication process are shown in Table 1 below.

[0137] Table 1. Predicted and measured data during biological doubling.

[0138] Time (h) MA Predicted Value MA measured value PR Predictions PR measured value 0 0.62 0.65 0.48 0.50 12 0.70 0.73 0.51 0.53 24 0.81 0.83 0.58 0.61 36 0.87 0.88 0.62 0.63 48 0.90 0.89 0.64 0.60 60 0.92 0.91 0.74 0.73 72 0.94 0.93 0.77 0.76

[0139] According to Table 1 and Figure 5 The data graph showing the process of biological multiplication illustrates that, throughout the entire process of biological multiplication, Figure 5 MA shown in (a) and Figure 5 (b) shows that the model prediction and measured values ​​of PR change over time in a highly consistent and similar manner, indicating that the established model has high prediction accuracy and good fitting effect, and can accurately characterize the dynamic changes in microbial activity and the proportion of dominant strains.

[0140] The cultivated highly active microorganisms were added to the biological treatment tank at multiple points using a uniform dosing method (initial dosing rate of 100 L / h). Samples of the influent, effluent, and mixed liquor within the tank were collected periodically to monitor COD. Ammonia nitrogen, total phosphorus (TP), microbial concentration, and microbial activity parameters were integrated and standardized to construct a wastewater treatment status dataset. This dataset was then input into an improved random forest wastewater treatment status prediction model. This model, after stratified sampling, Bayesian hyperparameter optimization, and time-series weighted fusion, outputs the water quality change trend for the next one cycle (15 minutes). Based on the prediction bias, a three-dimensional linkage hierarchical control was initiated: primary control adjusted the microbial inoculation rate; secondary control adjusted the aeration intensity of the biological treatment tank (base value 500 m³ / h); and tertiary control supplemented nutrients to ensure water quality stability. The wastewater treatment process data were statistically analyzed, resulting in the predicted and measured data in Table 2 below.

[0141] Table 2. Predicted and measured data during wastewater treatment.

[0142] time 0h 1h 2h 3h 4h 5h 5.5h COD Predicted Value 410.2 mg / L 290.5 mg / L 200.3 mg / L 140.7 mg / L 90.2 mg / L 45.6 mg / L 27.8 mg / L COD measured value 412.7 mg / L 289.1 mg / L 202.2 mg / L 139.5 mg / L 91.8 mg / L 44.8 mg / L 28.5 mg / L <![CDATA[BOD5 predicted value]]> 185.4 mg / L 120.1 mg / L 75.2 mg / L 50.5 mg / L 32.4 mg / L 18.3 mg / L 12.3 mg / L <![CDATA[Measured value of BOD5]]> 187.1 mg / L 121.5 mg / L 74.5 mg / L 51.2 mg / L 31.9 mg / L 18.9 mg / L 12.1 mg / L Ammonia nitrogen prediction value 52.1 mg / L 34.2 mg / L 20.5 mg / L 11.3 mg / L 5.8 mg / L 2.6 mg / L 1.1 mg / L Measured ammonia nitrogen value 52.9 mg / L 33.9 mg / L 20.9 mg / L 11.5 mg / L 6.0 mg / L 2.4 mg / L 1.2 mg / L TP Predicted Value 6.3 mg / L 4.1 mg / L 2.6 mg / L 1.6 mg / L 0.9 mg / L 0.5 mg / L 0.4 mg / L TP measured value 6.5 mg / L 4.0 mg / L 2.8 mg / L 1.5 mg / L 1.0 mg / L 0.6 mg / L 0.5 mg / L

[0143] According to Table 2 and Figure 6 The wastewater treatment process data chart shown indicates that throughout the entire wastewater treatment process, Figure 6 (a) COD, Figure 6 (b) , Figure 6 (c) ammonia nitrogen, Figure 6 (d) The predicted and measured values ​​of TP over time are highly consistent with each other with minimal deviation, indicating that the constructed model has high prediction accuracy, strong generalization and reliability, and can accurately track the dynamic changes of pollutant degradation.

[0144] The effluent indicators were measured for three consecutive cycles (45 min) and found to meet the Class A discharge standard for urban wastewater treatment plants (COD≤50mg / L, ammonia nitrogen≤5mg / L, TP≤0.5mg / L). When the model predicted that the water quality would consistently meet the standards in the future, the effluent discharge procedure was initiated. At this time, the measured water quality data were: COD 27mg / L, ammonia nitrogen 1.1mg / L, TP 0.35mg / L, microbial activity MA 0.93, and dominant strain percentage PR 0.77.

[0145] This embodiment details the process of improving the efficiency of wastewater biochemical treatment using this method in an industrial park integrated wastewater treatment plant as an application scenario. During the process, the invention accurately characterizes the dynamic changes in microbial activity and the proportion of dominant strains through an improved LSTM biomultiplication state prediction model, providing solid data support for the precise simulation, reliable prediction, and stable control of the biomultiplication process. Furthermore, the improved random forest wastewater treatment state prediction model accurately tracks the dynamic changes in pollutant degradation, providing a reliable basis for the precise prediction, intelligent control, and efficient operation of the wastewater treatment process, thus achieving a deep improvement in the efficiency of wastewater biochemical treatment.

[0146] Example 3, based on Example 1 or 2, describes in detail a comparative experiment conducted under the experimental environment described in Example 1, using existing methods such as T-BDM (conventional biomultiplication method), LSTM-RF (LSTM-random forest water quality prediction method), I-BDM (integrated biomultiplication device method), BDM-MBR (biomultiplication coupled combined aeration MBR method), and ML-OPT (machine learning process optimization method) in batches, combined with the experimental results of this method, as follows:

[0147] In existing comparative methods, T-BDM (Traditional Biological Multiplication Method) uses a conventional biological multiplication tank, introduces a commonly acclimatized microbial community, sets fixed operating parameters (temperature 28℃, aeration intensity 2.0m³ / h, fixed nutrient salt dosage), and monitors the effluent water quality regularly. When the water quality is found to be substandard, the operating parameters are adjusted.

[0148] LSTM-RF (LSTM-Random Forest Water Quality Prediction Method): This method uses conventional biological treatment pond technology, directly adds conventional bacterial agents to the biological treatment pond, continuously collects water quality data from the biological treatment pond and constructs a dataset, inputs the dataset into the LSTM-RF dual model, and adjusts the aeration intensity and nutrient addition amount of the biological treatment pond according to the water quality change trend output by the model.

[0149] I-BDM (Integrated Biomultiplication Device): An integrated biomultiplication wastewater treatment device that adopts a multi-stage biomultiplication tank with a mechanically layered structure. The device is equipped with basic sensors to monitor temperature and dissolved oxygen parameters during operation. After microbial communities are introduced into the device, it will continue to operate under preset fixed conditions.

[0150] BDM-MBR (Biomultiplication Coupled Aeration MBR): Construct a diversified biomultiplication coupled combined aeration MBR device, introduce microbial communities adapted to cold-region environments into the device, and increase the dissolved oxygen supply in the device through combined aeration processes to optimize the nitrogen and phosphorus removal effect under low-temperature conditions.

[0151] ML-OPT (Machine Learning Process Optimization): Using conventional biological treatment tank technology, conventional bacterial agents are added to the biological treatment tank, and various parameters during the operation of the biological treatment tank are continuously collected. The collected data is analyzed through machine learning algorithms to optimize the aeration intensity, nutrient salt dosage and other operating parameters of the biological treatment tank.

[0152] Each method used wastewater with the same composition as in Example 1 (with slight deviations after artificial preparation), and operated continuously in batches for 30 days. During this period, wastewater was treated and discharged before a new batch of wastewater was introduced. Relevant experimental data were recorded, and the overall performance comparison data of each method was obtained in Table 3 below:

[0153] Table 3. Overall performance comparison data of each method

[0154] detection indicators T-BDM LSTM-RF I-BDM BDM-MBR ML-OPT This method Microbial activity MA 0.71 0.78 0.83 0.86 0.79 0.92 Percentage of dominant strains (PR) 0.61 0.67 0.63 0.69 0.71 0.76 effluent COD 42.3 mg / L 35.5 mg / L 40.2 mg / L 31.6 mg / L 36.9 mg / L 28.8 mg / L Ammonia nitrogen in effluent 3.8 mg / L 2.1 mg / L 3.5 mg / L 1.6 mg / L 2.2 mg / L 1.2 mg / L Water output TP 0.85 mg / L 0.52 mg / L 0.79 mg / L 0.49 mg / L 0.50 mg / L 0.38 mg / L Outgoing water meets standards and stability 76.7% 86.7% 80.0% 93.3% 86.7% 98.6% Wastewater treatment compliance time 8.2h 7.1h 7.7h 6.3h 6.0h 5.5h Time required to complete biological multiplication 84.2h 77.5h 80.3h 72.6h 69.4h 67.8h

[0155] Based on Table 3 and Figure 7 The comparison chart of the biomultiplication process data for each method shows that the biomultiplication process completed by this method takes the shortest time, and Figure 7 (a) shows that this method is significantly higher than the other 5 control methods in terms of microbial activity, demonstrating a stronger microbial activation ability; Figure 7 (b) shows that this method performs better in terms of the proportion of dominant strains, demonstrating a higher enrichment efficiency of dominant bacterial groups.

[0156] Based on Table 3 and Figure 8 The comparison chart of wastewater treatment process data for each method shows that this method has the shortest wastewater treatment time to meet standards, and Figure 8 (a) shows that this method produces the lowest COD content in the effluent. Figure 8 (b) shows that this method produces the lowest ammonia nitrogen content in the effluent. Figure 8 (c) shows that the effluent TP concentration of this method is the lowest, demonstrating stronger pollutant removal capacity and higher treatment efficiency, with better effluent quality and faster system compliance.

[0157] Based on Table 3 and Figure 9 As shown in the comprehensive performance comparison chart of the various methods, this method has the highest microbial activity, the proportion of dominant strains, and the stability of effluent compliance. At the same time, it has the lowest effluent COD, ammonia nitrogen, TP concentration, wastewater treatment compliance time, and biological multiplication time.

[0158] This embodiment describes in detail a comparative experiment conducted in the experimental environment described in Example 1, using five typical methods from existing methods in batches, combined with the experimental results of this method. The experiment verifies that the method of this invention, through the dynamic regulation of the improved LSTM biological multiplication state prediction model, exhibits superior microbial activation and microbial community enrichment capabilities; through the improved random forest wastewater treatment state prediction model, it exhibits stronger pollutant degradation efficiency; and based on the synergistic effect of the above models, it achieves more stable effluent quality and a shorter treatment and multiplication cycle.

[0159] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication, characterized in that, include: Intelligent biological multiplication is carried out by building a stratified microbial activation multiplier and introducing an initial population of dominant microorganisms adapted to the wastewater to be treated. The integrated sensor built into the hierarchical microbial activation multiplier collects data from the multiplier. Based on the collected data, the microbial activity and community ratio are calculated, and a biological multiplication status dataset is constructed by combining the collected data. Based on the biomultiplication state dataset, an improved LSTM biomultiplication state prediction model is used to predict the changing trends of microbial activity and community ratio within the multiplier. Based on this trend, the parameters of the multiplier were dynamically adjusted to obtain highly active microorganisms. Wastewater is treated biologically by adding highly active microorganisms to the biological treatment tank and conducting regular water quality monitoring to obtain a wastewater treatment status dataset. Based on the wastewater treatment status dataset, an improved random forest wastewater treatment status prediction model is used to predict the trend of water quality changes in the biochemical pond. Based on the trend of water quality changes, the microbial inoculation rate and the operating parameters of the biological treatment tank are dynamically adjusted. When the indicators meet the preset emission standards for multiple consecutive tests and the model predicts that the water quality will remain stable and meet the standards in the future, the effluent discharge procedure will be initiated to complete the biological treatment of the wastewater.

2. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication as described in claim 1, characterized in that, The layered microbial activation multiplier is specifically divided into an upper activation chamber, a middle proliferation chamber, and a lower maturation chamber from top to bottom. The orderly flow of the microbial community is achieved between the chambers through an annular inclined flow guide component. Each chamber of the multiplier is independently equipped with integrated sensors, including a temperature sensor, a dissolved oxygen sensor, a nutrient concentration sensor, and an ultrasonic sensor. The multiplier also integrates a constant temperature heating module, a micro-nano aeration module, and a nutrient slow release module to provide a controllable environment for microbial activation and proliferation.

3. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication as described in claim 1, characterized in that, The biomultiplication state dataset is specifically defined as follows: by integrating sensors, key parameters of each chamber of the multiplier are collected at a preset frequency, including at least the temperature, dissolved oxygen concentration, nutrient concentration and microbial concentration of each chamber, and the bioactivity and microbial community ratio are calculated; wherein the bioactivity is calculated based on the microbial concentration collected by the ultrasonic sensor and the dehydrogenase activity detection method is used; the biomultiplication state dataset is constructed by combining the collected data and the calculated data.

4. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication as described in claim 1, characterized in that, The improved LSTM biomultiplication state prediction model specifically involves: inputting a biomultiplication state dataset, using an improved LSTM algorithm to output prediction results at a preset frequency, including predicted values ​​of microbial activity and the proportion of dominant strains in each chamber; outputting the prediction results to an intelligent control terminal, comparing them with preset target thresholds, and generating targeted dynamic control decisions for multiplier parameters based on the predicted trends.

5. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 4, characterized in that, The improved LSTM algorithm specifically involves: weighting the hidden layer outputs at different time points using an attention mechanism, assigning higher weights to parameters that significantly affect microbial activity and bacterial community proportions, as shown in the formula: in, for Attention weight at any moment To score attention, , These are the attention weight matrix and the attention bias term, respectively. for Predicted output at time step The time series length of the biological doubling state dataset. For time steps, for The hidden layer output at each moment; Using a genetic algorithm, a random combination of hyperparameters is generated as the initial population of individuals. The fitness function is the model prediction error, and individual selection is performed using a roulette wheel selection method, with the following formula: in, For model prediction error, To predict the sample size, For the first The actual value of each sample For the first The predicted value for each sample, For the fitness function, Choose the probability for the individual. Population size; A single-point crossover method is adopted, randomly selecting a crossover point and exchanging partial gene fragments of two parent individuals to generate offspring individuals. The mutation method of the offspring individuals is to randomly select individual gene positions for flipping. When the number of iterations of the genetic algorithm reaches a preset value or the fitness function value tends to stabilize, the hyperparameter combination corresponding to the individual with the highest fitness is output. Based on this hyperparameter combination, the improved LSTM is iteratively optimized.

6. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 5, characterized in that, The dynamic control decision-making is specifically as follows: based on the preset high activity standard threshold, microbial community ratio adaptation standard threshold, and synergistic effect standard threshold, combined with the predicted value, the current multiplier operating parameters are maintained when the predicted value meets the target threshold, and the data is continuously monitored. When the predicted value does not meet the target threshold, the parameter adjustment mechanism is activated to adjust the operating parameters of constant temperature heating, aeration intensity, and nutrient supply.

7. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 1, characterized in that, The wastewater treatment status dataset specifically includes: the addition of highly active microorganisms using a multi-point uniform addition method, with the addition points evenly distributed in the upper, middle, and lower layers of the biological treatment tank; regular collection of influent, effluent, and mixed liquor samples from the biological treatment tank; and monitoring parameters including chemical oxygen demand (COD) and five-day biochemical oxygen demand (BOD5). ammonia nitrogen Total phosphorus (TP) and microbial concentration Microbial activity (MA); integrate all monitoring data and construct a wastewater treatment status dataset in a three-dimensional structure of time-monitoring parameters-reaction conditions.

8. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 1, characterized in that, The improved random forest wastewater treatment status prediction model specifically involves: inputting a wastewater treatment status dataset, applying an improved random forest algorithm, and outputting future water quality change trends, including COD, etc. , The predicted value of TP; when the deviation between the predicted water quality parameter and the target value is greater than the target water quality threshold, the dynamic adjustment of water quality treatment parameters is initiated. Based on the water quality prediction trend, the corresponding parameters are adjusted to ensure that the water quality is maintained within the target range.

9. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 1, characterized in that, The improved random forest algorithm specifically involves sampling the wastewater treatment status dataset using a stratified sampling method, assigning higher sampling weights to low-frequency water quality anomaly samples. A dual-objective optimization function is constructed to optimize both water quality prediction accuracy and error convergence speed. The optimal hyperparameter combination is iteratively sought using a Bayesian optimization algorithm for four core hyperparameters: number of decision trees, tree depth, minimum number of split samples, and minimum number of leaf node samples. The formula for the dual-objective optimization function is as follows: in, Fitness values ​​optimized for Bayesian approaches. Weighting coefficients for water quality prediction accuracy optimized by Bayes. For the number of test samples, , The first The predicted value and the actual value of each test sample. This represents the error convergence rate. Based on the optimal hyperparameter combination, a forest of multiple decision trees is constructed. The weighted Gini coefficient is used as the decision tree splitting rule, assigning higher splitting weights to key features affecting water quality prediction accuracy. The formula is as follows: in, This is the weighted Gini coefficient; the smaller the value, the higher the node purity and the better the splitting effect. The first step is to set the degree of influence of characteristics on water quality prediction. Weighting factors for water quality characteristics For the number of water quality parameter categories, For the first Under class characteristics, the first The sample proportion of water quality parameters of the same type; In the multi-decision-tree fusion stage, a time-series weighting factor is used to assign higher voting weights to the prediction results of the decision trees corresponding to recent wastewater treatment status data. The prediction results of all decision trees are then weighted and fused to obtain the predicted value of the target water quality parameter, as shown in the formula: in, These are predicted values ​​for water quality parameters. For the number of decision trees, For the first The time-series weighting factor of the time step. For the first The decision tree in the th Predicted values ​​at time steps The total length of the time series. For time steps.

10. The method for improving the efficiency of wastewater biochemical treatment based on intelligent biological multiplication according to claim 1, characterized in that, The dynamic control of water quality treatment parameters specifically involves: establishing a three-dimensional linkage system of microbial dosing, biochemical tank operation, and nutrient replenishment; classifying control priorities based on prediction deviations; and performing control operations in stages according to control priorities. Among them, the primary regulation adjusts the rate of addition of highly active microorganisms, the proportion of addition points and the addition cycle, prioritizing the rapid improvement of water quality through the action of microorganisms; Secondary regulation adjusts the operating parameters of the biochemical tank to optimize the microbial living environment; The three-level regulation linkage activates the precise nutrient replenishment module, which calculates the deficiencies in carbon, nitrogen, and phosphorus sources based on prediction deviations and provides targeted supplementation of nutrients suitable for the growth of microbial communities.

Images