Sewage treatment process optimization control system based on ASM2 and BP models
By combining ASM2 and BP models, an optimized control system for wastewater treatment was constructed, which solved the problems of sludge generation and reduction in water quality and energy consumption control. This achieved stable effluent from the wastewater treatment plant, optimized energy consumption, and reduced sludge volume, thereby improving operational stability and refined control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-03-27
AI Technical Summary
Existing research focuses primarily on optimizing water quality parameters and controlling energy consumption, with less consideration given to sludge generation and reduction control. This leads to excessive sludge production, increased dewatering and disposal costs, and environmental risks.
By combining the ASM2 mechanism model with the BP neural network, an optimized control system for the wastewater treatment process is constructed to achieve joint optimization control of effluent quality, energy consumption, and sludge production. Through data acquisition, mechanism modeling, prediction, and control modules, the aeration rate, return ratio, and sludge discharge rate are dynamically adjusted.
It has achieved stable effluent quality that meets standards from wastewater treatment plants, reduced energy consumption and sludge production, achieved the goal of energy-saving and efficient sludge reduction, and improved operational stability and refined control.
Smart Images
Figure CN121742383A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automated control technology for wastewater treatment and environmental engineering, specifically to an optimized control system for wastewater treatment processes based on the ASM2 and BP models. Background Technology
[0002] In the operation of modern wastewater treatment plants, the activated sludge system is the core component for pollutant removal, and its operational stability directly affects effluent quality and system energy consumption. Traditional operation control methods mainly rely on manual experience to adjust process parameters such as aeration rate, return ratio, and sludge age, which makes it difficult to achieve precise control in complex dynamic environments, leading to problems such as fluctuating effluent quality, high energy consumption, and excessive sludge production.
[0003] The Activated Sludge Model (ASM) is an internationally widely used mechanism model for wastewater biochemical processes, which can comprehensively describe processes such as organic matter degradation, nitrification, denitrification, and biological phosphorus removal. Among them, the ASM2 dynamic mechanism model better simulates the synergistic removal process of carbon, nitrogen, and phosphorus by introducing reaction characteristic parameters of energy storage substances (such as polyphosphate-accumulating bacteria).
[0004] Meanwhile, machine learning techniques, especially back propagation neural networks (BP), have demonstrated excellent performance in nonlinear system modeling and dynamic prediction. BP models learn the complex mapping relationship between system inputs and outputs based on historical data, exhibiting good generalization and adaptive capabilities. Their typical structure includes an input layer, hidden layers, and an output layer, adjusting weights and biases through an error backpropagation algorithm to minimize prediction errors. In wastewater treatment, BP neural networks have been used to predict effluent quality indicators such as COD and ammonia nitrogen, but these applications are mostly limited to optimizing single water quality parameters and lack comprehensive control over the sludge formation process.
[0005] Existing research largely focuses on optimizing water quality parameters and controlling energy consumption, with less consideration given to sludge generation and reduction. Excessive sludge production not only increases dewatering and disposal costs but also poses potential environmental risks. Therefore, there is an urgent need to establish an intelligent control system that integrates mechanistic and data-driven models to achieve dynamic optimization of sludge reduction while ensuring stable effluent quality compliance. Summary of the Invention
[0006] The technical problem this invention aims to solve is that existing research mainly focuses on the optimization of water quality parameters and energy consumption control, with less consideration given to sludge generation and reduction regulation. Excessive sludge production not only increases dewatering and disposal costs but also poses potential environmental risks. The goal is to provide an optimized control system for wastewater treatment processes based on the ASM2 and BP models. By coupling the ASM2 mechanistic model with the BP neural network prediction model, joint optimization control of effluent quality, energy consumption, and sludge production is achieved, thereby realizing energy-saving, high-efficiency, and sludge-reduced operation. This system improves the accuracy of effluent quality prediction, ensures stable compliance, dynamically optimizes aeration volume, return ratio, and sludge discharge strategies to reduce energy consumption, and establishes a sludge production prediction model to achieve sludge reduction.
[0007] This invention is achieved through the following technical solution:
[0008] This invention provides a wastewater treatment process optimization control system based on the ASM2 and BP models, comprising:
[0009] The data acquisition module is used to collect key process parameters of wastewater treatment in real time.
[0010] The ASM2 mechanism modeling module is used to construct ASM2 dynamic mechanism models to simulate organic matter degradation, nitrification and denitrification, biological phosphorus removal and sludge generation and decay processes, and to determine key state variables.
[0011] The BP neural network prediction module is used to obtain joint predictions of effluent quality and sludge production based on key state variables.
[0012] The sludge reduction prediction and control module is used to dynamically generate sludge reduction control strategies based on joint prediction values and the sludge generation mechanism of the ASM2 dynamic mechanism model.
[0013] The optimized control module aims to achieve multiple objectives, including stable effluent quality, minimal sludge production, and minimum operating energy consumption. It coordinates the optimization of operating parameters such as aeration rate, return ratio, and sludge discharge rate, and sends optimization instructions to the actuators.
[0014] Furthermore, the data acquisition module includes:
[0015] High-precision sensors are used to collect parameters such as influent chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, water volume, dissolved oxygen, pH value, temperature, and sludge concentration in real time.
[0016] Signal conditioning circuitry is used to filter and amplify the acquired signal;
[0017] The data communication unit is used to transmit the data processed by the signal conditioning circuit to the central processing unit, and to perform abnormal data processing and normalization processing.
[0018] Furthermore, the ASM2 mechanism modeling module includes:
[0019] A dynamic mechanism model of ASM2 was constructed based on the biochemical reaction kinetic equations in ASM2 to simulate the processes of organic matter degradation, nitrification and denitrification, biological phosphorus removal, and sludge generation and decay.
[0020] Historical operational data is input into the ASM2 dynamic mechanism model, and a genetic algorithm is used to calibrate the ASM2 dynamic mechanism model to obtain the calibrated ASM2 dynamic mechanism model.
[0021] Key state variables were calculated using the calibrated ASM2 dynamic mechanism model.
[0022] Furthermore, the key state variables include the biomass concentrations of heterotrophic bacteria, autotrophic bacteria, and polyphosphate-accumulating bacteria.
[0023] Furthermore, the BP neural network prediction module includes:
[0024] The input layer acquires microbial concentration variables from the output of the ASM2 dynamic mechanism model and real-time monitoring data from the data acquisition module.
[0025] The output layer outputs predicted values for chemical oxygen demand (COD), total nitrogen (TNO), total phosphorus (TP) concentrations, and sludge production in the effluent.
[0026] Furthermore, the sludge reduction prediction and control module includes:
[0027] Based on the sludge generation kinetic equation of the ASM2 dynamic mechanism model and the sludge production prediction value output by the BP neural network, a sludge production prediction equation is established.
[0028] The sludge reduction index (SRI) is introduced as an evaluation indicator to assess sludge reduction.
[0029] The discharge rate and aeration intensity of excess sludge are dynamically adjusted based on sludge reduction.
[0030] Further, sludge production prediction involves the following specific processes:
[0031] Real-time collection of influent flow rate Water inlet concentration, Concentration, sludge age Dissolved oxygen and water temperature ;
[0032] The parameters collected in real time are input into the trained BP neural network model to obtain the correction factor used to compensate for the unmodeled dynamics.
[0033] The correction factor used to compensate for unmodeled dynamics and the parameters collected in real time are combined and substituted into the sludge production prediction equation:
[0034] ;
[0035] The predicted value of sludge production per unit time was obtained.
[0036] Furthermore, the Sludge Reduction Index (SRI) specifically includes:
[0037] Obtain the predicted value of sludge production per unit time
[0038] Obtain the baseline sludge production rate under the same operating conditions.
[0039] Calculate the sludge reduction index: ;
[0040] Sludge reduction based on the sludge reduction index evaluation system.
[0041] Furthermore, the optimization control module includes:
[0042] Obtain predicted effluent indicators, effluent setpoints, predicted sludge production, and aeration energy consumption;
[0043] A multi-objective optimization function is constructed with the following comprehensive objectives: stable effluent quality compliance threshold, minimum sludge production, and optimal energy consumption threshold.
[0044] Aeration rate, recirculation ratio, and sludge discharge ratio were selected as optimization variables. A multi-objective optimization algorithm was used to solve the objective function to obtain the optimal aeration rate, optimal recirculation ratio, and optimal sludge discharge ratio.
[0045] Based on the optimal values of the optimal aeration rate, optimal reflux ratio, and optimal sludge discharge ratio, the aeration rate, internal and external reflux ratio, and the excess sludge discharge ratio are dynamically adjusted.
[0046] Furthermore, the construction of the multi-objective optimization model includes:
[0047] ;
[0048] in, To predict water discharge indicators, Set the output water value. For aeration energy consumption, These are the weighting coefficients.
[0049] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0050] The integration of mechanism and data-driven approaches: By using the ASM2 dynamic mechanism model to provide physical constraints and the BP model for dynamic correction, the limitations of a single model are overcome, and the accuracy of system prediction and control is improved.
[0051] Joint optimization control: The system can simultaneously predict and control effluent quality and sludge production, realizing intelligent and refined operation of the wastewater treatment plant, and achieving significant sludge reduction.
[0052] Energy consumption and cost optimization: By coordinating the operation of aeration volume, reflux ratio and sludge discharge volume, energy consumption and chemical dosage are effectively reduced, and equipment life is extended.
[0053] Adaptive and learning capabilities: The BP model supports online learning and can adapt to complex operating conditions such as fluctuations in influent load and temperature changes, and has long-term operational stability.
[0054] Scalability and extensibility: The system features a modular design, making it flexible for application in municipal wastewater treatment plants, industrial park wastewater treatment and reclaimed water systems, and it supports integration with other intelligent algorithms. Attached Figure Description
[0055] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:
[0056] Figure 1 This is a system overall structure diagram in an embodiment of the present invention. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0058] As one possible implementation method, such as Figure 1As shown, this embodiment provides a wastewater treatment process optimization control system based on the ASM2 and BP models. This system comprehensively utilizes the ASM2 (Activated Sludge Model No. 2) dynamic mechanism model and the BP (Back Propagation) neural network surrogate model to perform real-time prediction and optimization control of key operating parameters in the wastewater biochemical treatment process. The system includes a data acquisition module, a model calculation module, an optimization decision-making module, and a control execution module. The data acquisition module acquires real-time data on influent water quality such as COD, BOD, ammonia nitrogen, total nitrogen, total phosphorus, water volume, dissolved oxygen (DO), pH, temperature, sludge concentration (MLSS), and sludge age (SRT) through high-precision sensors. The model calculation module simulates the reaction kinetics process based on the ASM2 dynamic mechanism model and learns and corrects the system output through the BP model. The optimization decision-making module generates control strategies by combining multi-objective optimization algorithms. The control execution module feeds back optimization commands to units such as aeration rate, return ratio, and sludge discharge rate to achieve dynamic adaptive control. This embodiment, through the deep integration of mechanism and data-driven approaches, can effectively improve the operational stability and energy efficiency of wastewater treatment plants, achieving the comprehensive optimization of pollutant removal rate, operating cost, and sludge reduction. It is applicable to municipal wastewater treatment plants, industrial park wastewater treatment, and reclaimed water systems.
[0059] The specific implementation process includes:
[0060] Step 1: Data Acquisition and Preprocessing
[0061] The system collects parameters such as influent flow rate, chemical oxygen demand (COD), ammonia nitrogen, total nitrogen (TNO), total phosphorus (TP), pH value, temperature, and MLSS using high-precision online monitoring instruments, with a sampling frequency of 1-5 minutes per sampling. The collected signals are filtered and amplified by a signal conditioning circuit. The industrial data communication unit transmits the processed data to the central processing unit for anomaly processing and normalization. After synchronization, the data is stored in a real-time database for model retrieval.
[0062] Step 2: Simulation and Dynamic Calculation of ASM2 Dynamic Mechanism Model
[0063] A mechanistic model is constructed based on the biochemical reaction kinetic equations in ASM2 to simulate the processes of organic matter degradation, nitrification / denitrification, biological phosphorus removal, and sludge formation and decay. The ASM2 dynamic mechanistic model uses matrix notation to describe these processes, where components and transformation processes are represented by subscripts i and j, respectively, and the mass of substances produced and consumed should be equal in each transformation process. Key model parameters are calibrated using a genetic algorithm based on historical operating data to calculate key state variables such as the biomass concentrations of heterotrophic bacteria, autotrophic bacteria, and polyphosphate-accumulating bacteria. The model simulates the removal of organic matter, ammonia nitrogen, total nitrogen, and total phosphorus from the influent, outputting predicted effluent water quality values and microbial concentration changes. The model runs for 1-2 hours and is updated synchronously with real-time data.
[0064] The matrix symbol expression is as follows: ,in This represents the production rate of component i in all parallel processes. This represents the stoichiometric coefficient of substance i in the j-th reaction. The vector representing the process rate equation.
[0065] Step 3: BP Neural Network Prediction and Error Correction
[0066] The input layer nodes of the BP neural network prediction module receive microbial concentration variables output from the ASM2 dynamic mechanism model and real-time monitoring data from the data acquisition module; the output layer nodes correspond to the predicted values of effluent COD, TN, TP concentrations, and sludge production. The hidden layer of the network uses the Sigmoid activation function, and the training process uses the mean squared error as the loss function, with iterative updates to the weights. The training data covers typical operating conditions, and after initial offline training, it transitions to online learning, triggering model correction based on prediction errors, and supporting online incremental learning.
[0067] The Sigmoid activation function is used, and the mean squared error is used as the loss function during training. Weights are iteratively updated, and the specific process includes:
[0068] ;
[0069] Where, Sigmoid(z) = , where W is the weight matrix and b is the bias vector;
[0070] The formula for mean square error is: Where n represents the number of samples, Represents the true value of the sample. This represents the predicted value of the sample.
[0071] Step 4: Sludge Reduction Prediction and Feedback Control
[0072] A time-varying sludge production prediction equation is established based on the sludge formation kinetic equation of the ASM2 dynamic mechanism model and the sludge production prediction value output by the BP neural network. By introducing the sludge reduction index (SRI) as an evaluation indicator, it is linked with the optimization control module to dynamically adjust the excess sludge discharge rate and aeration intensity (DO setpoint 0.5-2.0 mg / L). When the predicted sludge production exceeds the threshold, the SRT is automatically adjusted or the anoxic period duration is increased to promote sludge decay and achieve sludge production minimization control.
[0073] Sludge production prediction equation: ,in, This is the predicted sludge production per unit time. This is a correction factor output by the BP model, used to compensate for unmodeled dynamics.
[0074] Sludge Reduction Index (SRI) as an evaluation indicator: ,in This is the baseline sludge production.
[0075] Step 5: Multi-objective optimization and decision execution
[0076] The optimized control module constructs a multi-objective optimization model based on effluent quality index deviations, predicted sludge production, and aeration energy consumption functions. A particle swarm optimization algorithm is used to solve for the optimal solution set, dynamically adjusting the aeration rate, internal / external recirculation ratio (30%-100%), and excess sludge discharge ratio. Control commands are sent via PLC to the aerator (variable frequency control), recirculation pump, and sludge discharge system, achieving adaptive closed-loop control. The control cycle is 15-30 minutes to ensure timely system response.
[0077] ;
[0078] in, To predict water discharge indicators, Set the output water value. For aeration energy consumption, These are the weighting coefficients.
[0079] The optimization variables include aeration rate (controlled by a frequency converter to control the blower), return ratio (controlled by the return pump), and sludge discharge ratio (controlled by the sludge discharge valve).
[0080] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A wastewater treatment process optimization control system based on ASM2 and BP models, characterized in that, include: The data acquisition module is used to collect key process parameters of wastewater treatment in real time. The ASM2 mechanism modeling module is used to construct ASM2 dynamic mechanism models to simulate organic matter degradation, nitrification and denitrification, biological phosphorus removal and sludge generation and decay processes, and to determine key state variables. The BP neural network prediction module is used to obtain joint predictions of effluent quality and sludge production based on key state variables. The sludge reduction prediction and control module is used to dynamically generate sludge reduction control strategies based on joint prediction values and the sludge generation mechanism of the ASM2 dynamic mechanism model. The optimized control module aims to achieve multiple objectives, including stable effluent quality, minimal sludge production, and minimum operating energy consumption. It coordinates the optimization of operating parameters such as aeration rate, return ratio, and sludge discharge rate, and sends optimization instructions to the actuators.
2. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 1, characterized in that, The data acquisition module includes: High-precision sensors are used to collect parameters such as influent chemical oxygen demand, ammonia nitrogen, total nitrogen, total phosphorus, water volume, dissolved oxygen, pH value, temperature, and sludge concentration in real time. Signal conditioning circuitry is used to filter and amplify the acquired signal; The data communication unit is used to transmit the data processed by the signal conditioning circuit to the central processing unit, and to perform abnormal data processing and normalization processing.
3. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 1, characterized in that, The ASM2 mechanism modeling module includes: A dynamic mechanism model of ASM2 was constructed based on the biochemical reaction kinetic equations in ASM2 to simulate the processes of organic matter degradation, nitrification and denitrification, biological phosphorus removal, and sludge generation and decay. Historical operational data is input into the ASM2 dynamic mechanism model, and a genetic algorithm is used to calibrate the ASM2 dynamic mechanism model to obtain the calibrated ASM2 dynamic mechanism model. Key state variables were calculated using the calibrated ASM2 dynamic mechanism model.
4. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 3, characterized in that, The key state variables include the biomass concentrations of heterotrophic bacteria, autotrophic bacteria, and polyphosphate-accumulating bacteria.
5. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 1, characterized in that, The BP neural network prediction module includes: The input layer acquires microbial concentration variables from the output of the ASM2 dynamic mechanism model and real-time monitoring data from the data acquisition module. The output layer outputs predicted values for chemical oxygen demand (COD), total nitrogen (TNO), total phosphorus (TP) concentrations, and sludge production in the effluent.
6. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 1, characterized in that, The sludge reduction prediction and control module includes: Based on the sludge generation kinetic equation of the ASM2 dynamic mechanism model and the sludge production prediction value output by the BP neural network, a sludge production prediction equation is established. The sludge reduction index (SRI) is introduced as an evaluation indicator to assess sludge reduction. The discharge rate and aeration intensity of excess sludge are dynamically adjusted based on sludge reduction.
7. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 6, characterized in that, Sludge production forecasting involves the following process: Real-time collection of influent flow rate , water inlet concentration, Concentration, sludge age Dissolved oxygen and water temperature ; The parameters collected in real time are input into the trained BP neural network model to obtain the correction factor used to compensate for the unmodeled dynamics. The correction factor used to compensate for unmodeled dynamics and the parameters collected in real time are combined and substituted into the sludge production prediction equation: ; The predicted value of sludge production per unit time was obtained.
8. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 7, characterized in that, The Sludge Reduction Index (SRI) specifically includes: Obtain the predicted value of sludge production per unit time Obtain the baseline sludge production rate under the same operating conditions. Calculate the sludge reduction index ; Sludge reduction based on the sludge reduction index evaluation system.
9. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 1, characterized in that, The optimization control module includes: Obtain predicted effluent indicators, effluent setpoints, predicted sludge production, and aeration energy consumption; A multi-objective optimization function is constructed with the following comprehensive objectives: stable effluent quality compliance threshold, minimum sludge production, and optimal energy consumption threshold. Aeration rate, recirculation ratio, and sludge discharge ratio were selected as optimization variables. A multi-objective optimization algorithm was used to solve the objective function to obtain the optimal aeration rate, optimal recirculation ratio, and optimal sludge discharge ratio. Based on the optimal values of the optimal aeration rate, optimal reflux ratio, and optimal sludge discharge ratio, the aeration rate, internal and external reflux ratio, and the excess sludge discharge ratio are dynamically adjusted.
10. The wastewater treatment process optimization control system based on ASM2 and BP models according to claim 9, characterized in that, The construction of the multi-objective optimization model includes: ； in, To predict water discharge indicators, Set the output water value. For aeration energy consumption, These are the weighting coefficients.