Online calibration method for activated sludge model parameters based on adaptive genetic algorithm
The adaptive genetic algorithm-based online calibration method improves activated sludge model accuracy and optimizes wastewater treatment processes by adjusting parameters in real-time, reducing energy consumption.
Patent Information
- Authority / Receiving Office
- HK · HK
- Patent Type
- Applications
- Current Assignee / Owner
- BEIJING CHUANGCHUANG CHINA-DUTCH TECHNOLOGY CO LTD
- Filing Date
- 2026-05-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing wastewater treatment systems face challenges in maintaining accurate simulation and optimization of activated sludge models due to the lack of effective online calibration methods, leading to inefficiencies and increased energy consumption.
An online calibration method using an adaptive genetic algorithm is employed to automatically adjust activated sludge model parameters based on real-time data, first calibrating high-sensitivity parameters and then all parameters if necessary, ensuring the model meets operating conditions.
The method enhances simulation accuracy and supports process optimization while reducing energy consumption in wastewater treatment plants.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
Abstract The present application relates to the field of wastewater treatment technology, and in particular to an online calibration method for activated sludge model parameters based on an adaptive genetic algorithm. The method includes: A100, determining whether to automatically trigger an online calibration procedure for parameters of an activated sludge model according to a real-time data set of a wastewater treatment system; A200, if the online calibration procedure is automatically triggered, performing online calibration on the parameters of the activated sludge model by adopting a parameter grouping update strategy and an adaptive genetic algorithm; wherein, first calibrating high-sensitivity parameters among the parameters according to the real-time data set; judging whether the activated sludge model based on the calibrated high-sensitivity parameters satisfies model operating conditions; and if not, calibrating all parameters according to the real-time data set, and judging whether the activated sludge model based on all calibrated parameters satisfies the model operating conditions. The present application continuously ensures the simulation accuracy of the model, and provides reliable technical support for process optimization, energy saving and consumption reduction of wastewater treatment plants.
Claims
1. A method for online calibration of activated sludge model parameters based on an adaptive genetic algorithm, characterized in that, include: A100. Based on the real-time dataset of the wastewater treatment system, determine whether to automatically trigger the online calibration procedure for the parameters in the activated sludge model; The real-time dataset includes key parameters such as influent parameters, operational status parameters, and effluent water quality parameters. A200. If the online calibration procedure is automatically triggered, the parameters of the activated sludge model are calibrated online using an adaptive genetic algorithm according to the parameter grouping update strategy. The online calibration and the operation of the activated sludge model are performed in parallel. Specifically, based on the real-time dataset, the high-sensitivity parameters in the parameters are first calibrated; and it is determined whether the activated sludge model based on the calibrated high-sensitivity parameters meets the model operation conditions. If not, calibrate all parameters based on the real-time dataset and determine whether the activated sludge model based on all calibrated parameters meets the model operating conditions. If satisfied, the calibration parameters are stored and A100 is repeated.
2. The method according to claim 1, characterized in that, The A100 includes: A101. Collect the influent parameters, status operation parameters and effluent water quality parameters of the sewage treatment system in real time according to the specified data acquisition frequency, and form a standardized real-time dataset. A102. Based on the real-time dataset, the pre-determined multi-objective variables and weight information, obtain the final accuracy of the currently running activated sludge model, and based on the final accuracy, determine whether to automatically trigger the online calibration procedure for the parameters in the activated sludge model. The accuracy calculation of the activated sludge model is as follows: ； Where Fit represents the model error, used to characterize the model accuracy; the smaller the Fit value, the higher the model accuracy. i Let X be the weight of the i-th target variable. i,sim Let X be the simulated value of the i-th objective variable. i,meas Let be the measured value of the i-th target variable.
3. The method according to claim 2, characterized in that, Influent parameters include: flow rate, COD, ammonia nitrogen, total nitrogen, total phosphorus, SS, pH value and / or temperature; Operating parameters include: dissolved oxygen, sludge concentration, dosage, reflux ratio, and / or sludge discharge rate; The effluent water quality parameters include: COD, ammonia nitrogen, total nitrogen, total phosphorus and / or SS; The A101 includes: performing noise filtering, default value interpolation, and outlier identification using a sliding window method on the influent parameters, operating status parameters, and effluent water quality parameters to generate a standardized dataset.
4. The method according to claim 2, characterized in that, The A102 includes: If the final accuracy is greater than the preset threshold every time within a consecutive n time period, the online calibration program will be automatically triggered; or, if the final accuracy is greater than the preset threshold every p consecutive times, the online calibration program will be automatically triggered.
5. The method according to claim 2, characterized in that, In A200, determining whether an activated sludge model based on calibrated high-sensitivity parameters meets the model operating conditions includes: Based on the real-time dataset, obtain the new final accuracy of the activated sludge model using calibrated high-sensitivity parameters, and determine whether the new final accuracy is greater than the preset threshold at least twice. If it is, the model operation conditions are not met; otherwise, the model operation conditions are met. If not, calibrate all parameters based on the real-time dataset, and determine whether the activated sludge model based on all calibrated parameters meets the model operating conditions, including: If the new final accuracy of the activated sludge model using all calibrated parameters is greater than the preset threshold, then the model operation conditions are not met. The system automatically adjusts the crossover and mutation rates of the adaptive genetic algorithm in each iteration, recalibrates all parameters of the activated sludge model, obtains the new final accuracy of the activated sludge model using all calibrated parameters, and determines whether the activated sludge model based on all calibrated parameters meets the model operating conditions. If it does not meet the conditions, an alert is issued. or, Determine whether the activated sludge model based on the calibrated high-sensitivity parameters meets the model operating conditions, including: Using real-time datasets, the new final accuracy of the activated sludge model with calibrated high-sensitivity parameters is obtained. It is determined at least twice whether the new final accuracy is greater than a preset threshold. If it is, the model operation conditions are not met; otherwise, the model operation conditions are met. If the conditions are not met, the crossover rate and mutation rate of the adaptive genetic algorithm will be automatically adjusted in each iteration based on the real-time dataset to calibrate all parameters of the activated sludge model. The system will then determine whether the activated sludge model based on all calibrated parameters meets the model operation conditions. If not, an alert will be issued.
6. The method according to claim 1, characterized in that, In A200, an adaptive genetic algorithm is used to calibrate the parameters of the activated sludge model online, including: The parameters of the activated sludge model are calibrated online using real-number encoding or Gray encoding; and The online calibration process is performed using a multi-core GPU server, and the response time for a single calibration is less than 30 minutes.
7. The method according to claim 1, characterized in that, An adaptive genetic algorithm is used to calibrate the parameters of the activated sludge model online, including: Centered on the parameters currently used in the activated sludge model, multiple individuals are randomly generated within the value range corresponding to each parameter to form an initial population. Each individual in the initial population corresponds to a set of activated sludge model parameters. Based on the initial population, iterative optimization is performed with the goal of achieving the best model accuracy to obtain the optimal individual. This optimal individual corresponds to a set of calibrated activated sludge model parameters, which are used to replace the parameters of the current activated sludge model to complete the online calibration.
8. The method according to claim 7, characterized in that, The iterative optimization includes: The top 60% of individuals in terms of fitness are selected from the current population to enter the next generation. The fitness of an individual is determined based on the model accuracy calculated from the parameters of the activated sludge model corresponding to that individual. The better the model accuracy, the higher the fitness of the individual. The selected individuals undergo real-valued single-point crossover, and the crossover rate is dynamically adjusted based on the current evolutionary state of the population. Specifically, ； Among them, C p Let f be the crossover rate of the current individual, and f be the fitness of the current individual. avg f represents the average fitness of the current population. max This represents the maximum fitness of the current population. Gaussian mutation is performed on the individuals after crossover. The mutation rate of the individuals is dynamically adjusted according to the current population state, and the mutation amplitude is dynamically adjusted according to the parameter grouping. Specifically, ， Among them, M p Let f be the mutation rate of the current individual, and f be the fitness of the current individual. avg f represents the average fitness of the current population. min This represents the minimum fitness of the current population. A new generation of population is generated based on the results of selection, crossover, and mutation, and the above iterative process is repeated until the preset termination condition is met. The individual with the best fitness in the final population is taken as the optimal individual.
9. The method according to claim 3, characterized in that, For a given wastewater treatment plant, the pre-determined multi-objective variables include: Indicators from effluent water quality parameters include COD, ammonia nitrogen, total nitrogen, and total phosphorus at the secondary sedimentation tank effluent outlet; In addition, process indicators from the operating parameters include ammonia nitrogen, dissolved oxygen, and sludge concentration at the end of the aerobic tank, nitrate nitrogen and phosphate at the end of the anoxic tank, and phosphate at the end of the anaerobic tank.
10. The method according to claim 1, characterized in that, For a specific wastewater treatment plant, the high-sensitivity parameters include: autotrophic bacteria yield coefficient, heterotrophic bacteria yield coefficient, maximum specific growth rate of heterotrophic bacteria, maximum specific growth rate of autotrophic bacteria, substrate half-saturation constant, heterotrophic bacteria decay coefficient, autotrophic bacteria decay coefficient, and hydrolysis rate constant. Low-sensitivity parameters are other activated sludge model parameters besides high-sensitivity parameters.