CN-122021996-A - Multi-agent collaborative dosing optimization method and system for sewage plant

CN122021996ACN 122021996 ACN122021996 ACN 122021996ACN-122021996-A

Abstract

The invention discloses a multi-agent collaborative dosing optimization method and a system thereof for a sewage plant, belonging to the technical field of intelligent control of sewage treatment, wherein the method comprises a multi-source water quality data acquisition step, a pretreatment step and a water quality data acquisition step, wherein the water quality data of each process node is obtained; the method comprises a carbon source demand prediction step and a phosphorus removal agent demand prediction step, wherein an LSTM model integrating an attention mechanism is adopted to predict the addition amount, a multi-agent collaborative decision step is adopted to construct collaborative constraints based on the influence relationship of a carbon source on biological phosphorus removal efficiency, a multi-objective optimization problem which aims at agent cost, standard risk and carbon footprint is solved, a fuzzy evaluation correction step is adopted to dynamically correct the addition amount, a distributed addition execution step is adopted to determine the distribution proportion and the addition time sequence of each addition point according to hydraulic retention time, and the method realizes collaborative optimization of the carbon source and the phosphorus removal agent, improves the accuracy and the robustness of addition control, simultaneously brings the carbon footprint into an optimization objective and supports realization of carbon neutralization objective in a sewage plant.

Inventors

JIN JIE
YAO LING
DU YIBO
Wu Xuanbei
LEI YUNHUI
CHEN BOYU
LI JIANG

Assignees

四川发展国润水务投资有限公司
农业农村部成都沼气科学研究所

Dates

Publication Date: 20260512
Application Date: 20251222

Claims (10)

1. The multi-agent collaborative dosing optimization method for the sewage plant is characterized by comprising the following steps of: A multi-source water quality data acquisition step, namely acquiring water quality monitoring data of each process node of sewage treatment, and performing data cleaning and normalization pretreatment on the acquired water quality monitoring data to generate a standardized water quality characteristic sequence; A carbon source demand prediction step, namely inputting a standardized water quality characteristic sequence into a pre-trained carbon source addition prediction model, and outputting a carbon source addition prediction value under the current working condition based on the total nitrogen concentration of inlet water, the chemical oxygen demand of inlet water, the carbon-nitrogen ratio and the hydraulic retention time; A dephosphorization agent demand prediction step, namely inputting a standardized water quality characteristic sequence into a pre-trained dephosphorization agent dosage prediction model, and outputting a dephosphorization agent dosage prediction value under the current working condition based on the total phosphorus concentration of the inlet water, the phosphate concentration and the total phosphorus target value of the outlet water; A multi-agent collaborative decision-making step of receiving a predicted value of the carbon source addition amount and a predicted value of the dephosphorization agent addition amount, constructing a multi-agent collaborative constraint condition based on the influence relationship of the carbon source addition on the biological dephosphorization efficiency, and generating a collaborative addition strategy by solving a multi-objective optimization problem, wherein the multi-objective optimization problem takes the minimization of the total cost of the agent and the maximization of the effluent standard rate as optimization targets; A fuzzy evaluation correction step, namely performing rationality evaluation on the collaborative dosing strategy by adopting a fuzzy comprehensive evaluation method according to the deviation of the recent actual dosing effect and the expected effect, generating a dosing amount correction coefficient based on an evaluation result, and acting the correction coefficient on the collaborative dosing strategy to generate a corrected dosing instruction; And a distributed dosing execution step, wherein the distribution proportion and the dosing time sequence of the corrected dosing instruction at each dosing point are determined according to the process position and the hydraulic retention time of each dosing point, and the distributed dosing instruction is sent to an execution mechanism of the corresponding dosing point.
2. The method for optimizing multi-agent collaborative dosing in a sewage plant according to claim 1, wherein in the carbon source demand prediction step, carbon source dosing is triggered when the influent carbon nitrogen ratio is lower than a preset carbon nitrogen ratio threshold, the carbon source dosing prediction model is a long-term and short-term memory network model integrating an attention mechanism, and the model takes recent influent water quality time sequence data and process operation parameters as input characteristics.
3. The method for optimizing multi-agent collaborative dosing of a sewage plant according to claim 1, wherein in the step of predicting the demand of the dephosphorizing agent, the dosing of the dephosphorizing agent is triggered when the phosphate concentration of the effluent is higher than a preset phosphate concentration threshold value, and the feature extraction layer is shared by a dephosphorizing agent dosing amount prediction model and a carbon source dosing amount prediction model.
4. The method for optimizing multi-agent co-addition in a sewage plant according to claim 1, wherein in the multi-agent co-decision step, the co-constraint conditions include an upper limit constraint of carbon source addition, an upper limit constraint of dephosphorization agent addition, a total nitrogen concentration constraint of effluent, a total phosphorus concentration constraint of effluent, and an incremental constraint of chemical oxygen demand caused by carbon source addition.
5. The method for optimizing multi-agent cooperative dosing of a sewage plant according to claim 1, wherein in the step of fuzzy evaluation and correction, a multi-level evaluation index system comprising the fluctuation degree of inflow water quality, the predicted deviation amplitude and the effluent standard reaching margin is established, the weight of each evaluation index is determined by adopting a analytic hierarchy process, and the value of the dosing quantity correction coefficient is determined by fuzzy synthesis operation.
6. The method for optimizing multi-agent cooperative dosing of a sewage plant according to claim 1, wherein in the step of performing distributed dosing, carbon source dosing points comprise an anoxic tank and a deep denitrification filter tank, dephosphorization agent dosing points comprise an anoxic tank and a secondary lifting pump station tank, and the distribution proportion of each dosing point is dynamically adjusted according to denitrification and dephosphorization loads of each tank.
7. The method for optimizing multi-agent co-dosing in a wastewater treatment plant of claim 1, wherein the multi-objective optimization problem further comprises a carbon footprint cost term calculated from the product of the production carbon emission intensity and the dosing amount of the carbon source agent and the dephosphorizing agent.
8. The multi-agent collaborative dosing optimization method for the sewage plant according to claim 1, wherein when the continuous preset times of the evaluation result of the fuzzy evaluation correction step is in an unreasonable interval, an incremental learning process of a carbon source dosing prediction model and a phosphorus removal agent dosing prediction model is triggered.
9. The method for optimizing multi-agent cooperative dosing of a sewage plant according to claim 1, wherein the step of executing distributed dosing further comprises determining a dosing execution time corresponding to current water inflow according to hydraulic retention time between a dosing point and a water inlet, and enabling the dosed agent and the corresponding water inflow to be converged in a target process pool.
10. Multi-agent collaborative dosing optimization system for sewage plants, which is characterized by comprising: The multi-source water quality data acquisition module is used for acquiring water quality monitoring data of each process node of sewage treatment, and performing data cleaning and normalization pretreatment on the acquired water quality monitoring data to generate a standardized water quality characteristic sequence; The carbon source demand prediction module is used for inputting the standardized water quality characteristic sequence into a pre-trained carbon source addition prediction model and outputting a carbon source addition prediction value based on the total nitrogen concentration of the inflow water, the chemical oxygen demand of the inflow water, the carbon nitrogen ratio and the hydraulic retention time; The dephosphorization agent demand prediction module is used for inputting the standardized water quality characteristic sequence into a pre-trained dephosphorization agent dosage prediction model and outputting a dephosphorization agent dosage prediction value based on the total phosphorus concentration of the inflow water, the phosphate concentration and the total phosphorus target value of the outflow water; the multi-agent collaborative decision-making module is used for receiving the predicted value of the carbon source addition amount and the predicted value of the phosphorus removal agent addition amount, constructing a collaborative constraint condition based on the influence relationship of the carbon source addition to the biological phosphorus removal efficiency, and generating a collaborative addition strategy by solving a multi-objective optimization problem aiming at minimizing the total cost of the agent and maximizing the yielding water standard reaching rate; the fuzzy evaluation correction module is used for reasonably evaluating the cooperative dosing strategy by adopting a fuzzy comprehensive evaluation method, generating a dosing amount correction coefficient based on an evaluation result, and applying the correction coefficient to the cooperative dosing strategy to generate a corrected dosing instruction; The distributed dosing execution module is used for determining the distribution proportion and the dosing time sequence of the corrected dosing instruction according to the process position and the hydraulic retention time of each dosing point, and sending the distributed dosing instruction to the execution mechanism of the corresponding dosing point.

Description

Multi-agent collaborative dosing optimization method and system for sewage plant Technical Field The invention belongs to the technical field of intelligent control of sewage treatment, and particularly relates to a multi-agent collaborative dosing optimization method and system for a sewage plant. Background The urban domestic sewage treatment facility is one of the modern urban infrastructures, and with the acceleration of urban process, the domestic sewage discharge is greatly increased, and a serious challenge is provided for the efficient, low-cost and standard treatment of urban domestic sewage. The A2/O technology is most widely applied to urban domestic sewage treatment, and utilizes the nitrification and denitrification of microorganisms in an activated sludge process to degrade harmful substances in sewage. Due to the water quality characteristics of low carbon and high nitrogen and phosphorus of domestic sewage, under the increasingly strict discharge standard requirements, a sewage treatment plant has to be supplemented with a certain carbon source and a phosphorus removing agent to ensure that the effluent reaches the standard. Chinese patent CN120802857a discloses a control system for a sewage treatment plant, the system comprising a multi-modal biosensing module, a cloud edge cooperative data processing module and an adaptive decoupling controller. The cloud edge cooperative data processing module comprises an edge computing module and a cloud optimization platform, wherein the cloud optimization platform operates a multi-target dynamic optimization module and a digital twin model. The multi-target dynamic optimization module in the prior art synchronously optimizes targets of reaching quality standards, minimizing energy consumption and reducing chemical agents based on pretreatment data, and performs optimization calculation by adopting an NSGA-III algorithm. However, the above prior art has a technical problem that, first, the technique mainly focuses on the optimized addition of a single agent (carbon source), and does not sufficiently consider the synergistic relationship between the carbon source and the dephosphorizing agent. In the actual sewage treatment process, the adding of the carbon source can influence the biological dephosphorization efficiency, because the phosphorus accumulating bacteria needs volatile fatty acid as the carbon source to perform anaerobic phosphorus release, and the carbon source can be competitively consumed by other microorganisms due to the improper opportunity and dosage of the added carbon source, thereby reducing the biological dephosphorization efficiency and increasing the dosage of the chemical dephosphorization agent. Secondly, the optimization decision of the technology lacks a real-time evaluation and dynamic correction mechanism aiming at the adding effect, and when the quality of the inflow water greatly fluctuates, the accuracy of the adding amount is difficult to ensure only by means of a prediction model. In addition, this technique does not take into account the indirect carbon emission problem of dosing of agents, and lacks the ability to incorporate the carbon footprint into an optimization objective in the current dual carbon objective context. Disclosure of Invention The invention aims to provide a multi-agent collaborative dosing optimization method and a multi-agent collaborative dosing optimization system for a sewage plant, so as to solve the technical problems that in the prior art, the carbon source and the dephosphorization agent are lack of collaborative optimization, the dosing effect is lack of real-time correction, and the carbon footprint cost is not considered. The invention provides a multi-agent collaborative dosing optimization method for a sewage plant, which comprises a multi-source water quality data acquisition step, a data processing step and a data processing step, wherein the multi-source water quality data acquisition step is used for acquiring water quality monitoring data of all process nodes of sewage treatment, and performing data cleaning and normalization pretreatment on the acquired water quality monitoring data to generate a standardized water quality characteristic sequence; the method comprises a step of predicting the carbon source demand, a step of predicting the standardized water quality characteristic sequence, a step of receiving a carbon source dosing amount prediction value and a dephosphorization agent dosing amount prediction value based on the total inlet nitrogen concentration, the chemical oxygen demand of inlet water, the carbon nitrogen ratio and the hydraulic retention time, a step of outputting a carbon source dosing amount prediction value under the current working condition, a step of predicting the dephosphorization agent demand, a step of inputting the standardized water quality characteristic sequence into the pre-trained dephosphorization agent dosing amount prediction model, a ste