A method and device for treating MRS sewage based on multi-source data and a storage medium

Abstract

The present disclosure provides a multi-source data-based MRS sewage treatment method, device and storage medium. The method comprises: acquiring a total set of data associated with sewage at a plurality of monitoring nodes in a target area; constructing a graph in a space-time dimension based on the total set of data; processing the graph using a trained first network model to determine a route segment with a fault risk; and processing the graph and the route segment with the fault risk using a trained second network model to timely adjust the combination of MRS water purification equipment to be invested. The method significantly improves the efficiency of the entire MRS sewage treatment system, reduces energy consumption, and overall improves the intelligence and accuracy of sewage treatment, providing efficient technical support for rural water environment governance.

Description

Technical Field

[0001] This disclosure generally relates to the field of MRS wastewater treatment, and particularly to an MRS wastewater treatment method, apparatus and storage medium based on multi-source data. Background Technology

[0002] In recent years, MRS water purification technology (microbial high-efficiency water purification technology) has been gradually applied in rural areas. MRS water purification technology utilizes microbial communities (such as aerobic bacteria and anaerobic bacteria) to form a multi-layered bioreactor system, decomposing pollutants through microbial metabolism. Simultaneously, it combines plant-assisted purification to achieve highly efficient decomposition (see existing technology patent CN202110884371.0), exhibiting significant advantages such as low energy consumption, easy maintenance, and high environmental protection standards. However, MRS water purification technology also has its drawbacks. Microbial decomposition is slow, resulting in a long recovery period after the initial water intake cycle. Furthermore, the bacterial community in each tank is difficult to quantify, making it hard to accurately determine the recovery period. In most cases, assessments rely on experience, leading to significant discrepancies between the recovery period and the actual value. Additionally, with rural economic development and improved living standards, rural domestic sewage discharge is increasing annually, requiring the deployment of multiple MRS water purification units for sewage treatment. However, rural domestic sewage treatment suffers from dispersed distribution, management difficulties, and outdated monitoring methods. Most rural areas rely on manual, periodic inspections, making it difficult to monitor sewage discharge and treatment in real time. Data from each monitoring node is isolated, hindering unified management. Moreover, the purification cycle (i.e., recovery cycle) of each MRS water purification unit differs. Therefore, scientifically scheduling multiple MRS water purification units (i.e., adjusting the combination of deployed MRS water purification units in real time based on the overall sewage treatment system and the specific conditions of each unit) is also a challenge. Summary of the Invention

[0003] To address the above problems, in a first aspect of this disclosure, a method for MRS wastewater treatment based on multi-source data is provided, comprising: acquiring a total set of data related to wastewater at multiple monitoring nodes in a target area, wherein the total set of data includes at least a wastewater production data set, a wastewater pipe network data set, an MRS wastewater treatment data set, and an environmental data set; constructing a spatiotemporal graph based on the total set of data, wherein the graph includes a set of graph nodes and a set of graph edges, wherein the graph edges indicate the spatial relationships between sewage outlets, pipe networks, and the inlet and outlet of MRS water purification equipment; processing the graph using a trained first network model to determine route segments with fault risk; and processing the graph and the route segments with fault risk using a trained second network model to adjust the combination of MRS water purification equipment to be deployed in a timely manner, wherein the second network model includes a scheduling network model and an MRS cycle prediction network model, wherein the scheduling network model is used to adjust the combination of deployed MRS water purification equipment in a timely manner according to the actual situation of the entire wastewater treatment system, and the MRS cycle prediction network model is used to accurately predict the recovery cycle of each MRS water purification device and transmit the data to the scheduling network model.

[0004] In a second aspect of this disclosure, an MRS wastewater treatment device based on multi-source data is provided for performing the method described in the first aspect, comprising: a data acquisition module configured to acquire a total set of data associated with wastewater at multiple monitoring nodes in a target area, the total set of data including at least a wastewater production data set, a wastewater pipe network data set, an MRS wastewater treatment data set, and an environmental data set; a map construction module configured to construct a spatiotemporal map based on the total set of data, the map including a set of graph nodes and a set of graph edges, wherein the graph edges indicate the spatial relationships between sewage outlets, pipe networks, and the inlet and outlet of MRS water purification equipment; a risk segment determination module configured to process the map using a trained first network model to determine route segments with fault risk; a network scheduling module configured to adjust the combination of deployed MRS water purification equipment in a timely manner according to the actual situation of the entire wastewater treatment system; and an MRS cycle prediction module configured to accurately predict the cycle of each MRS water purification device and transmit the data to the network scheduling module.

[0005] In a third aspect of this disclosure, a computer-readable storage medium is provided. A computer program is stored on the medium, which, when executed by a processor, implements the method of the first aspect.

[0006] This invention organically combines artificial intelligence technology with the MRS wastewater treatment system, which can not only accurately predict the recovery cycle of MRS water purification equipment, but also, based on this, adjust the combination of MRS water purification equipment in a timely manner according to the wastewater production situation and the wastewater pipe network situation. This significantly improves the efficiency of the entire MRS wastewater treatment system and reduces energy consumption, thereby enhancing the intelligence and precision of wastewater treatment as a whole and providing efficient technical support for rural water environment management. Attached Figure Description

[0007] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0008] Figure 1 A flowchart of a multi-source data-based MRS wastewater treatment method according to some embodiments of the present disclosure is shown;

[0009] Figure 2 Three dynamic parameters are shown. , , A graph showing the changes over a certain period of time;

[0010] Figure 3 The recovery cycle of the MRS water purification equipment is shown. Figure 2 The graph shows the changes in dynamic parameters over a given time period.

[0011] Figure 4 The architecture diagram of the MRS periodic prediction network model is shown.

[0012] Figure 5 A block diagram of a multi-source data-based MRS wastewater treatment apparatus according to some embodiments of the present disclosure is shown;

[0013] Figure 6 An electronic device in which one or more embodiments of the present disclosure may be implemented is shown. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should be noted that the headings of any section / subsection provided herein are not limiting. Various embodiments are described throughout this document, and embodiments of any type may be included under any section / subsection. Furthermore, embodiments described in any section / subsection may be combined in any way with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0016] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below. The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0017] The following description will continue with reference to the accompanying drawings, which will provide some exemplary embodiments of this disclosure.

[0018] In general, this embodiment provides an MRS wastewater treatment system based on multi-source data. The system may include a front-end monitoring module, a data transmission module, a cloud platform management module, and a terminal application module.

[0019] The front-end monitoring module includes smart sensors installed at multiple monitoring nodes, which acquire data from these nodes. The data transmission module sends the real-time data acquired by the front-end monitoring module to the cloud platform management module. The cloud platform management module manages and processes the collected data (historical and current data), sending the results to the terminal application module. The terminal application module displays the processing results (e.g., route segments with potential faults, optimal combinations of MRS water purification equipment) to the user. The terminal application module provides a convenient operating interface for management personnel. For example, the terminal application module can be a web-based platform suitable for global monitoring and data analysis in a fixed office location, or a mobile app for inspection personnel to view data and receive tasks on-site.

[0020] The following details the process of the MRS wastewater treatment method based on multi-source data. Figure 1 A flowchart of a multi-source data-based MRS wastewater treatment method according to some embodiments of the present disclosure is shown.

[0021] like Figure 1 As shown in box 110, the total set of data related to sewage at multiple monitoring nodes in the target area is obtained. The total set of data includes at least the sewage production data set, the sewage pipeline network data set, the MRS sewage treatment data set, and the environmental data set.

[0022] Monitoring nodes can include points at each stage of wastewater generation, collection, treatment, and discharge, such as household sewage outlets, MRS water purification equipment outlets, and multiple pre-set monitoring points in the pipeline network. Various sensors are installed at these monitoring nodes to collect key data in real time, including wastewater flow, water quality, pipeline network status, and MRS equipment operation. These sensors transmit the collected data to the front-end monitoring module. The front-end monitoring module aggregates and categorizes the received data into wastewater production data sets, wastewater pipeline network data sets, MRS wastewater treatment data sets, and environmental data sets. For example, flow sensors and pH sensors can be installed at household sewage outlets to monitor wastewater discharge volume and initial pH in real time. Online COD (Chemical Oxygen Demand) and ammonia nitrogen monitors can be installed at the MRS water purification equipment outlet to obtain COD, ammonia nitrogen, and phosphorus content values. Pressure sensors and flow sensors can be installed at multiple points in the pipeline network to obtain pressure and flow rates at upstream and downstream monitoring nodes. Various temperature sensors, humidity sensors, wind sensors, and rainfall sensors can be installed at multiple locations associated with wastewater in the target area to monitor environmental changes.

[0023] The collected data is categorized by the front-end monitoring module. In some embodiments, the wastewater production data set includes: wastewater flow rate, water quality parameters, and wastewater turbidity at the farmer's discharge outlet, wherein the water quality parameters include wastewater pH value, chemical oxygen demand (COD) value, ammonia nitrogen value, and phosphorus content; the MRS wastewater treatment data set includes: influent flow rate, temperature, pH, dissolved oxygen, influent COD, influent ammonia nitrogen, pollutant concentration, transmembrane pressure difference, influent cycle, recovery cycle, and distance between each MRS water purification device and the discharge outlet; the wastewater pipeline network data set includes: pipeline pressure value, pressure value and flow rate at multiple monitoring nodes upstream and downstream; the environmental data set includes: ambient temperature, ambient humidity, rainfall, and wind speed. It should be understood that the location of monitoring nodes, sensors, collected data, and their types mentioned above are only examples and can be determined according to the different human and geographical environments of the target area. This disclosure does not impose specific limitations in this regard.

[0024] In existing technologies, rural sewage treatment is mostly decentralized and relies on manual inspection, resulting in isolated data and an inability to achieve unified management. The embodiments disclosed in this paper construct a digital closed loop covering the entire process through end-to-end integration, merging and classifying scattered datasets, breaking down the data silos at each monitoring point, and facilitating subsequent unified data management and analysis.

[0025] Next, the data transmission module can utilize hybrid communication methods. For example, wired transmission can be used in areas with short distances and convenient wiring (such as around sewage treatment plants) to ensure data transmission stability. For scattered farm drainage outlets and remote pipeline nodes, low-power wireless communication technologies (such as NB-IoT) can be used to reduce deployment costs. Simultaneously, the data transmission module can encrypt the transmitted data to prevent data leakage. By combining the high stability of wired transmission with the low power consumption and low cost advantages of wireless transmission, and ensuring security through encryption, the stability of data transmission and system robustness are addressed.

[0026] The data transmission module can send large amounts of collected data to the cloud platform management module, which is the core of the system. In some embodiments, edge computing modules can be integrated into the sensors at the monitoring nodes, such as by adding a lightweight AI chip, to perform local data preprocessing (e.g., outlier filtering, trend analysis), uploading only key data to the cloud platform management module. For example, embedding edge algorithms in farmers' drainage outlet devices can identify abnormal flow in real time and trigger local alarms, reducing the transmission burden. This can reduce system latency and cost, and improve response speed.

[0027] The cloud platform management module is used for data storage, processing, analysis, and decision-making. In some embodiments, the cloud platform management module may include a data storage unit, a data processing and analysis unit, and an intelligent decision-making unit. The data storage unit can use a distributed database to meet the long-term storage needs of massive amounts of data. The data processing and analysis unit uses algorithms to remove abnormal data and integrates multi-source information to form a complete wastewater treatment data chain. The intelligent decision-making unit analyzes historical and real-time data to determine the optimal combination of each MRS water purification device for each time period, and automatically pushes early warning information, route segments with fault risk, and maintenance route suggestions when a fault is detected (e.g., abnormal pipeline pressure may be predicted as a blockage point). This is described in detail below.

[0028] The following describes the specific process by which the cloud platform management module processes the overall data set.

[0029] In box 120, a spatiotemporal graph is constructed based on the total data set. The graph includes a set of graph nodes and a set of graph edges, where the graph edges indicate the spatial relationships between sewage outlets, pipe networks, and the inlet and outlet of MRS water purification equipment.

[0030] The graph node set includes the feature vectors of all monitoring nodes mentioned above at time t. For example, the graph node set includes the wastewater flow vector i(t), the network pressure vector j(t), the wastewater turbidity vector k(t), the ambient temperature vector w(t), the wastewater flow vector q(t) at the inlet / outlet of the MRS water purification equipment, the water quality vector g(t) at the inlet / outlet of the MRS water purification equipment, and so on.

[0031] The graph edge set includes the connection relationships of the pipe network, the connection relationships of sewage outlets, and the connection relationships of the inlet and outlet of the MRS water purification equipment. Specifically, the connection relationships of the pipe network are determined based on factors such as pipe diameter, pipe length, material, and flow direction; the connection relationships from sewage outlet nodes to their downstream monitoring nodes or pipe segment nodes; the connection relationships from monitoring nodes in the pipe network to the inlet node of the MRS water purification equipment; and the connection relationships from the outlet node of the MRS water purification equipment to the monitoring nodes of the downstream pipe network, etc.

[0032] In some embodiments, a graph node set is determined based on historical and current data associated with wastewater at multiple monitoring nodes; a graph edge set is determined based on the connection relationships of the pipeline network, the connection relationships of the sewage outlets, and the connection relationships of the inlet and outlet of the MRS water purification equipment; and a graph neural network is used to process the graph node set and the graph edge set to determine the graph spectrum.

[0033] Graph neural networks preferably include heterogeneous graph neural networks, which include, but are not limited to, RGCN (Relational Graph Convolutional Network), HAN (Heterogeneous Graph Attention Network), and HGT (Heterogeneous Graph Transformer).

[0034] The specific processing procedure for using graph neural networks to process the graph node set and graph edge set to determine the graph spectrum is as follows: The feature vectors of all monitoring nodes in the graph node set at time t and all connection relationships in the graph edge set are input into a heterogeneous graph neural network (such as RGCN or HAN) to perform a round of spatial information aggregation, resulting in a node vector representation H(t) that integrates the spatial and temporal information of the entire "sewage outlet-pipeline network-MRS water purification equipment inlet / outlet". Then, the time series [H(tn),..., H(t-1),H(t)] is input into a time series model (such as GRU, LSTM or Transformer) to capture the dynamic evolution of the state and output the "graph" of the entire system at the current time t, that is, a system state vector containing rich spatiotemporal and semantic information.

[0035] This embodiment utilizes the graph output by the graph neural network to indicate the complex causal chain of sewage outlet A affecting monitoring point C through pipe B, and ultimately flowing into the MRS water purification equipment. This unifies the data from the sewage outlet, the pipe network, and the MRS water purification equipment into the same graph, enabling the model to understand the mutual influence of each element in the causal chain, thereby forming a global and strongly correlated graph.

[0036] In box 130, the trained first network model is used to process the graph and identify route segments with fault risk.

[0037] The first network model can be a spatiotemporal graph neural network model. This model is used to determine, based on the graph, the impact of anomalies at a monitoring node on one or more adjacent monitoring nodes, and how this impact evolves over time, thereby predicting route segments that may be at risk of failure.

[0038] In some embodiments, a graph neural network is trained using first training data, which includes feature data of various blockage types in the spatiotemporal dimension. The feature data is determined based on historical and current data associated with sewage at multiple monitoring nodes. The trained graph neural network is used to process the graph and determine the blockage type, which includes slow-deposition blockage and sudden foreign object blockage. Based on the graph and the blockage type, route segments with failure risk are identified.

[0039] Specifically, data related to various blockage types is extracted from historical data, such as abnormal data at monitoring nodes, the time and location of the anomalies, and adjacent locations affected by the blockage. This data is then categorized according to blockage type (slow deposition, sudden foreign objects), and the feature data of each category is labeled, resulting in a labeled feature dataset. This labeled feature dataset is used as the first training data to train a graph neural network, such as a GNN model.

[0040] A trained graph neural network is used to process the graph at time t, outputting probability values ​​for different congestion categories, such as P(normal), P(slow deposition), and P(sudden foreign object). For example, when the model detects a strong spatiotemporal abrupt change in the flow of a monitored node, with a sharp drop in flow within a very short period and a simultaneous surge in pressure at all its upstream nodes, it classifies it as a "sudden foreign object congestion." When the model detects a continuous decrease in the flow of a monitored node (e.g., a linear decrease of more than 20% over the past 24 hours) and a lagging increase in pressure, it classifies it as a "slow deposition congestion." When the model outputs a high-confidence congestion type (e.g., a 95% probability of "sudden foreign object congestion"), the analysis of the fault risk route segment is triggered. It should be understood that the model detects the changing patterns of the monitored nodes, and this disclosure does not impose specific limitations in this regard.

[0041] In some embodiments, a subgraph is determined from the map based on the congestion type. The subgraph includes multiple congestion points and multiple nodes located upstream of the congestion points. Data at each congestion point and multiple nodes located upstream of the congestion point in the subgraph is monitored within a preset time period. If the data changes at the congestion point and multiple nodes located upstream of the congestion point match the data changes corresponding to the congestion type, the congestion point and multiple nodes located upstream of the congestion point are identified as multiple candidate nodes. The route segment connected by the multiple candidate nodes is identified as a route segment with fault risk.

[0042] For example, tools such as GNN Explainers or attention mechanisms can be used to automatically identify the subgraph that contributes most to the congestion type; this subgraph is the "fault-risk route segment." As an example, when a trained graph neural network outputs a slow-deposition pattern with a 92% confidence level, the GNN Explainer or attention mechanism returns a slow-deposition subgraph, such as one including congestion point C and upstream nodes A and B. Data from nodes A, B, and C over the past 24 hours is obtained, and the "slow-deposition pattern" rule (i.e., the data change pattern at slow-deposition nodes, such as a continuous decrease in flow and an increase in pressure at the end node of the route segment) is used to determine the fault-risk route segment. If in a slowly depositing subgraph, the flow rate at node A > the flow rate at pipe AB > the flow rate at node B > the flow rate at pipe BC > the flow rate at node C, meaning the flow rates at nodes A, B, and C are continuously decreasing, and the pressure at node C has started to rise in the last 2 hours, then the data change patterns at nodes A, B, and C are highly consistent with the data change patterns of the "slowly depositing" type. In this case, the fault risk route segments are identified as route segments AB and BC.

[0043] Continue to refer to Figure 1 In box 140, a trained second network model is used to process the map and route segments with failure risks, thereby adjusting the combination of MRS water purification equipment that needs to be put into operation in a timely manner.

[0044] The second network model includes a scheduling network model and an MRS cycle prediction network model. The scheduling network model is used to adjust the combination of MRS water purification equipment deployed in a timely manner based on the actual situation of the entire wastewater treatment system (e.g., which MRS water purification equipment to deploy and which to shut down to bring the equipment into its recovery cycle). The MRS cycle prediction network model is used to accurately predict the recovery cycle of each MRS water purification equipment and transmit the data to the scheduling network model for real-time adjustment of the deployed MRS water purification equipment.

[0045] Current methods for predicting the lifecycle of microbial water purification systems have two major limitations: 1. Purely data-driven methods (such as traditional LSTM): While capable of fitting complex nonlinear relationships, their predictions lack clear biochemical mechanistic explanations and have poor interpretability. Extrapolation prediction capabilities may decline when system operating conditions change significantly (e.g., drastic changes in influent water quality or microbial community turnover), and they require a large amount of high-quality data for training. 2. Purely mechanistic model methods (such as the classic Monod equation): However, their key parameters (such as maximum specific growth rate)... half-saturation constant In real, unsteady-state systems, the periodicity is dynamic, not a fixed constant. Using fixed parameters determined in manuals or laboratories for field periodic prediction will lead to large errors and is difficult to apply.

[0046] The MRS cycle prediction network model of this invention deeply integrates a "data-driven" LSTM model with a "mechanism-driven" Monod equation. The LSTM is responsible for identifying and predicting the dynamic parameters in the mechanistic equation in real time and dynamically, while the improved Monod equation uses these dynamic parameters to calculate the water purification cycle based on explicit biochemical principles. This method combines the powerful fitting ability of artificial intelligence with the physical interpretability and extrapolation robustness of the mechanistic model.

[0047] The MRS cycle prediction network model of this invention is decomposed into two levels: The first level (dynamic parameter identification): Utilizing an LSTM network to learn historical operational data, it outputs current and future key parameters of microbial dynamics that vary with conditions. The second level (mechanistic cycle calculation): Substituting the dynamic parameters output by the LSTM into a recovery cycle calculation formula derived and modified from the Monod equation, it directly calculates the theoretical recovery cycle. .

[0048] The core formula for predicting the recovery cycle of MRS water purification equipment is:

[0049]

[0050] in, : The predicted remaining recovery period of the MRS water purification equipment within the time period t+1.

[0051] The initial contaminant concentration (mg / L) of the MRS water purification system is estimated by online sensors or preceding units.

[0052] The critical pollutant concentration (mg / L) at which the water quality of the MRS water purification equipment meets the standards is a preset threshold.

[0053] The half-saturation constant reflects the affinity of microorganisms for pollutants and is affected by factors such as temperature, pH, and toxic substances. Small size indicates strong affinity and the ability to grow relatively quickly even with low pollutant concentrations. Larger values ​​indicate weaker affinity, requiring higher pollutant concentrations to achieve higher rates.

[0054] This is the predicted half-saturation constant for the time interval t+1.

[0055] Dynamic maximum reaction rate ( = , and The parameters (defined below) reflect the overall metabolic activity of the microbial community.

[0056] This represents the predicted dynamic maximum reaction rate within the time interval t+1.

[0057] η: System efficiency correction coefficient (0 < η ≤ 1), a key LSTM learning parameter used to correct deviations between theoretical formulas and real-world complex systems (e.g., those with mass transfer limitations, dead zones, membrane fouling, etc.). η represents the actual operating efficiency of the system.

[0058] This is the efficiency correction coefficient for the predicted time period t+1.

[0059] Research has proven that , , Changes in these three parameters will affect the purification cycle. It has a significant effect; an increase in Ks will significantly prolong the duration. However, increasing Vmax will significantly shorten the time. The change in the efficiency correction coefficient η has an effect on It has a linear and symmetrical effect.

[0060] Instruction manual attached Figure 2 The dynamic variation curves of the half-saturation constant Ks, the dynamic maximum reaction rate Vmax, and the system efficiency correction coefficient η over a treatment period (0-50 hours) are shown. Each parameter exhibits periodic fluctuations, simulating the fluctuations in key kinetic parameters caused by changes in environmental conditions or microbial community activity during actual biological treatment processes.

[0061] Instruction manual attached Figure 3 The instructions attached are shown. Figure 2 The purification and recovery cycle is calculated from the dynamic parameter changes. The changes over time. The purple curve and filled area in the graph visually reflect the fluctuation range of the purification time required, while the red dashed line indicates the simulation period. The average value.

[0062] The principle behind the formula for predicting the recovery cycle of MRS water purification equipment is as follows:

[0063] 1. Establish the classical Monod equation:

[0064]

[0065] Where μ is the microbial specific growth rate, which is the rate at which unit biomass increases per unit time. It represents the speed of microbial growth and is affected by pollutant concentration, temperature, pH, etc. In the Monod model, it is mainly limited by S.

[0066] The maximum specific growth rate is the maximum specific growth rate that microorganisms can achieve when the pollutant concentration is very high. At this point, the growth of microorganisms is not limited by the pollutant concentration, but only by the upper limit of the microorganism's own physiological structure.

[0067] S represents the pollutant concentration, which is the concentration of nutrients that limit microbial growth in a culture or reaction system, such as carbon sources (organic matter), nitrogen sources, and phosphorus sources. In biological wastewater treatment, the concentration of biodegradable organic matter is often expressed as COD or BOD. COD (Chemical Oxygen Demand): The amount of oxygen (mg / L) consumed by the oxidation of organic matter in water under the action of strong oxidants (such as potassium dichromate). BOD (Biochemical Oxygen Demand): The amount of dissolved oxygen (mg / L) consumed by microorganisms to decompose organic matter in water under aerobic conditions.

[0068] It is the half-saturation constant. That is, when The concentration of pollutants at a given time reflects the affinity of microorganisms for the pollutants.

[0069] 2. Modification and Derivation of the Monod Equation: In an MRS water purification system, assuming a relatively stable microbial concentration, the pollutant removal rate v can be expressed as:

[0070]

[0071] Where S represents the pollutant concentration;

[0072] X represents the dynamic effective microbial concentration / activity, which, unlike traditional MLSS, refers more to the active components with degradation functions.

[0073] For the maximum specific pollutant consumption rate, and Related, and The microbial yield coefficient Y is defined as follows, which is related to the key parameter:

[0074]

[0075] Unit: g - biomass (dry weight) / g - substrate, where the substrate is the pollutant.

[0076] and The relationship between them is:

[0077] For the above differential equation, starting from the initial concentration (Initial concentration of influent / reactor) to threshold concentration Integrate the concentration of the treated water to obtain the theoretical water purification recovery period by solving for the time T. The calculation formula is as follows:

[0078]

[0079] Considering = Moreover, in actual engineering, the mass transfer term (C0-C) t The impact of ) is relatively minor, and to enhance the practicality and robustness of the model, a logarithmic term is used as the dominant term, and an efficiency coefficient η is introduced to comprehensively reflect the non-ideality of the system. This invention uses its core dominant term as the core calculation formula, thus obtaining

[0080]

[0081] 3. LSTM Coupling Methods

[0082] As per the instruction manual Figure 4 As shown, the architecture of the MRS periodic prediction network model of this invention comprises four main components:

[0083] (1) Input layer (blue rectangle): represents the model input interface, which receives observation data of 8 water quality characteristic parameters over 10 consecutive time steps;

[0084] The input is multivariate time series data of the system over the past T time windows:

[0085]

[0086]

[0087] (Including: influent flow rate, temperature, pH, dissolved oxygen, influent COD, influent ammonia nitrogen, pollutant concentration, transmembrane pressure difference, etc.)

[0088] (2) LSTM processing layer (green rectangle × 2), including:

[0089] The first LSTM processing layer (64 units) performs feature extraction and pattern recognition on the input time series data and has a time memory function. The second LSTM processing layer (32 units) further extracts high-level time series features and achieves feature dimensionality reduction.

[0090] (3) Fully connected layer (orange rectangle × 2), including:

[0091] The first fully connected layer (16 units): uses the ReLU activation function to perform a non-linear transformation on the LSTM output. The second fully connected layer (8 units): further fuses features to enhance the model's expressive power.

[0092] (4) Output layer (red rectangle): Outputs three target dynamic parameters , , It is used for the status assessment of water treatment processes.

[0093]

[0094] Here, θ represents all trainable weights of the LSTM network. t+1 represents the parameter prediction for a future time step (or a prediction step size).

[0095] The connecting arrows (gray) indicate the forward propagation direction of data between different layers of the neural network.

[0096] Finally, the dynamic parameters predicted by LSTM and the currently measured parameters are combined. Preset Substitute into the core prediction formula:

[0097]

[0098] Output This is the predicted remaining recovery period for the MRS water purification equipment.

[0099] In the formula of this invention , , Instead of being a fixed value determined through experiments, the water purification recovery cycle is a time-varying variable dynamically predicted by an LSTM neural network based on the real-time operating status of the system, which greatly improves the accuracy of the recovery cycle prediction.

[0100] The specific working steps of the MRS periodic prediction network model are as follows:

[0101] 1. Data preparation and preprocessing: Collect historical operating data and corresponding water purification recovery cycles for each MRS water purification device, and clean, align and normalize the data.

[0102] 2. LSTM network model training: including:

[0103] Step 2.1: Using historical time-series data The model is then subjected to supervised learning as input.

[0104] Step 2.2: Utilize historical data from the same time period , Based on the actual observed recovery period, the "equivalent" recovery time within that period can be calculated in reverse. , , The parameter values ​​(which can be obtained by transforming the formula and combining it with sliding window averaging or optimization algorithms; these techniques are well-known in the field and will not be elaborated further) serve as supervision labels for training the LSTM.

[0105] Step 2.3: Construct a multi-layer LSTM network, using mean squared error (MSE) as the loss function, and train the network to learn from... To these three dynamic parameters ( , , The mapping relationship between ).

[0106] 3. Online Deployment and Prediction:

[0107] Step 3.1: Deploy the trained LSTM model to the real-time system.

[0108] Step 3.2: The system collects data in real time, inputs it into the model, and obtains the predicted values ​​of dynamic parameters.

[0109] Step 3.3: Substitute the predicted values ​​of dynamic parameters into the recovery cycle prediction formula to calculate and update the predicted recovery cycle of the MRS water purification equipment in real time.

[0110] Step 3.4: When the recovery cycle of the MRS water purification equipment is lower than the warning threshold (i.e., the MRS equipment is about to complete the recovery cycle), an warning is triggered.

[0111] 4. Adaptive Model Update:

[0112] After each recovery cycle, the MRS water purification equipment adds the data from that cycle to the sample library and periodically fine-tunes and updates the LSTM model to adapt to long-term changes in the microbial community and process.

[0113] The MRS periodic prediction network model of the present invention has the following technical effects:

[0114] 1. Mechanism and data fusion for high prediction accuracy and interpretability: The prediction results have clear biochemical significance (through the Monod equation) and adaptability to complex real-world factors (through dynamic parameters and η learned by LSTM), with higher accuracy than single methods.

[0115] 2. Strong generalization and extrapolation capabilities: Even when the influent load or environmental conditions undergo changes that have not been experienced before, the model is less likely to produce absurd predictions due to the constraints of the core mechanism, making it more robust than a pure black box model.

[0116] 3. Achieve true online monitoring of dynamic parameters: This invention provides , , These key biochemical parameters, measured in soft form, can be used to monitor microbial activity and system health in real time, guiding process optimization.

[0117] 4. Reduced data requirements and lower engineering threshold: Compared to black-box models that require massive amounts of data to train, this method reduces data requirements and accelerates model convergence by introducing prior mechanism knowledge, making it easier to implement in practical engineering.

[0118] The scheduling network model in the second network model is a spatiotemporal graph neural network prediction model. The spatiotemporal graph neural network prediction model can be trained using the second training data, which includes historical time period maps, route segments with fault risks, and preset MRS water purification equipment combinations, to obtain the trained spatiotemporal graph neural network prediction model.

[0119] In some embodiments, the spatiotemporal graph neural network prediction model includes an optimizer and a predictor. The optimizer outputs an initial MRS water purification equipment combination; the following steps are executed at least once: the initial MRS water purification equipment combination is input to the predictor to obtain the operational status prediction result of the entire wastewater treatment system; the optimizer determines an objective function that satisfies the constraints based on the operational status prediction result, the constraints including at least minimum cost and water quality compliance; the optimizer updates the initial MRS water purification equipment combination according to the optimization algorithm; and the updated MRS water purification equipment combination is used as the new initial combination.

[0120] In some embodiments, the optimizer generates an initial MRS water purification equipment combination. These parameter combinations can be determined based on historical empirical values. The initial MRS water purification equipment combination, the current time t graph, and the route segments with fault risk are used as context information and input into the predictor to output the prediction results of the system operating status for the next N hours. For example, the prediction results of the operating status of each MRS water purification equipment may include graphs for future time periods (e.g., the next 1 hour, i.e., t+1, 6 hours, i.e., t+6), such as the flow and pressure changes of each node in the pipeline network, the changes in the water quality and quantity of the water entering the MRS water purification equipment (because faulty route segments will affect the water entering the equipment), and the predicted output of the water quality parameters at the outlet of the MRS water purification equipment under the preset MRS water purification equipment combination.

[0121] The optimizer constructs an objective function that satisfies the constraints based on the prediction results output by the predictor, making the prediction results closer to the true values, thereby filtering out predictions that deviate significantly from the true values. Because there are insurmountable physical constraints in wastewater treatment systems, constraint terms must be used to force the model to comply; otherwise, the prediction results are meaningless.

[0122] As an example, the first objective function satisfies the non-negativity constraint. By penalizing negative values ​​through the loss term, the predicted parameters, namely flow rate, concentration, and liquid level, are all non-negative. The first objective function is as follows:

[0123]

[0124] Where t∈{1,2,...,T} is the time step (e.g., one time step every 15 minutes, T is the total prediction duration), v is a single pipeline node (e.g., pump station, inspection well, inlet of MRS water purification equipment), V is the set of monitoring nodes, and N is the total number of pipeline nodes. These are the model predictions of flow rate / liquid level / COD concentration at node v at time t. If the value is less than 0, a penalty is imposed, forcing the model to predict a value greater than or equal to 0.

[0125] As an example, the second objective function satisfies the constraint that the prediction results conform to the law of conservation of water flow in the pipe network, that is, the total inflow flow = the total outflow flow, thus avoiding the contradiction of "inflow ≠ outflow". The second objective function is as follows:

[0126]

[0127] Where t∈{1,2,...,T} is the time step (e.g., one time step every 15 minutes, T is the total prediction duration), v is a single pipeline node (e.g., the inlet of a pumping station, inspection well, or MRS water purification equipment), Vmid is the set of intermediate nodes, in(v) is the set of upstream nodes of the inflow node v, and out(v) is the set of downstream nodes of the outflow node v. It is the predicted value of flow rate / liquid level / COD concentration of node u at time t. It is the predicted value of flow rate / liquid level / COD concentration at node w at time t.

[0128] The optimizer adjusts the initial MRS water purification equipment combination based on optimization algorithms (such as gradient descent and genetic algorithms) to obtain a second set of MRS water purification equipment combinations. This second set of combinations is then used as the new initial MRS water purification equipment combination, and the above steps are repeated. After hundreds or thousands of iterations, a set of MRS water purification equipment combinations with the optimal objective function (e.g., lowest cost while satisfying all constraints) is found. Finally, this optimal combination is sent to the PLC / DCS system of the MRS water purification equipment for real-time adjustments.

[0129] The complete process provided in this embodiment—including map construction, congestion type identification, fault risk section location, and combined optimization—achieves a closed-loop process from data perception to intelligent decision-making. Furthermore, it maximizes the efficiency of the entire wastewater treatment system while ensuring water quality meets standards, saving significant operation and maintenance costs for rural areas with relatively limited economic resources.

[0130] Figure 5 A schematic structural block diagram of a multi-source data-based MRS wastewater treatment device 200 according to some embodiments of the present disclosure is shown. The various modules / components in the device 200 can be implemented by hardware, software, firmware, or any combination thereof.

[0131] The device 200 includes a data acquisition module 210, configured to acquire a total set of data related to wastewater at multiple monitoring nodes in the target area. This data set includes at least a wastewater production data set, an MRS wastewater treatment data set, a wastewater pipe network data set, and an environmental data set. A graph construction module 220 is configured to construct a spatiotemporal graph based on the total data set. The graph includes a set of graph nodes and a set of graph edges, where the graph edges indicate the spatial relationships between sewage outlets, pipe networks, and the inlets and outlets of the MRS water purification equipment. A risk route determination module 230 is configured to use a trained first network model to process the graph and determine route segments with fault risks. A network scheduling module 240 is used to adjust the combination of deployed MRS water purification equipment in a timely manner according to the actual situation of the entire wastewater treatment system. An MRS cycle prediction module 250 is used to accurately predict the cycle of each MRS water purification device and transmit the data to the network scheduling module 240. The network scheduling module 240 uses a trained second network model to process the graph and route segments with fault risks, thereby adjusting the combination of MRS water purification equipment to be deployed in a timely manner. The MRS wastewater treatment device 200 is used to achieve the above and Figure 1 The methods and processes described herein.

[0132] Figure 6 A block diagram of an electronic device 300 in which one or more embodiments of the present disclosure may be implemented is shown. It should be understood that... Figure 6 The electronic device 300 shown is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. Figure 6 The electronic device 300 shown can be used to achieve Figure 1 Wastewater management methods.

[0133] like Figure 6 As shown, electronic device 300 is in the form of a general-purpose electronic device. Components of electronic device 300 may include, but are not limited to, one or more processors or processing units 310, memory 320, storage device 330, one or more communication units 340, one or more input devices 350, and one or more output devices 360. Processing unit 310 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 320. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 300.

[0134] Electronic device 300 typically includes multiple computer storage media. Such media can be any available media accessible to electronic device 300, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 320 can be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 330 can be a removable or non-removable medium and can include machine-readable media, such as flash drives, disks, or any other media that can be used to store information and / or data (e.g., training data for training) and can be accessed within electronic device 300.

[0135] Electronic device 300 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 6 As shown, disk drives for reading from or writing to removable, non-volatile disks (e.g., "floppy disks") and optical disk drives for reading from or writing to removable, non-volatile optical disks can be provided. In these cases, each drive can be connected to a bus (not shown) via one or more data media interfaces. Memory 320 may include computer program product 325 having one or more program modules configured to perform various methods or actions of various embodiments of this disclosure.

[0136] The communication unit 340 enables communication with other electronic devices via a communication medium. Additionally, the functionality of the components of the electronic device 300 can be implemented using a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, the electronic device 300 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.

[0137] Input device 350 can be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 360 ​​can be one or more output devices, such as a monitor, speaker, printer, etc. Electronic device 300 can also communicate with one or more external devices (not shown) via communication unit 340 as needed. These external devices include storage devices, display devices, etc., and can communicate with one or more devices that enable user interaction with electronic device 300, or with any device that enables electronic device 300 to communicate with one or more other electronic devices (e.g., network card, modem, etc.). Such communication can be performed via input / output (I / O) interface (not shown).

[0138] According to an exemplary implementation of this disclosure, a computer-readable storage medium is provided that stores computer-executable instructions thereon, wherein the computer-executable instructions are executed by a processor to implement the methods described above. According to an exemplary implementation of this disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, which are executed by a processor to implement the methods described above.

[0139] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0140] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0141] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0142] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0143] Various implementations of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is chosen to best explain the principles, practical applications, or improvements to technology in the market, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. A wastewater treatment method based on multi-source data using MRS, comprising: Obtain the total set of data related to sewage from multiple monitoring nodes in the target area. The total set of data includes at least the sewage production data set, the sewage pipeline network data set, the MRS sewage treatment data set, and the environmental data set. A spatiotemporal graph is constructed based on the total data set. The graph includes a set of graph nodes and a set of graph edges, where the graph edges indicate the spatial relationships between sewage outlets, pipe networks, and the inlet and outlet of MRS water purification equipment. The map is processed using a trained first network model to identify route segments with fault risk; the first network model is trained using first training data, which includes feature data of various congestion types in the spatiotemporal dimension. A trained second network model is used to process the graph and route segments with failure risks, thereby adjusting the combination of MRS water purification equipment to be deployed in a timely manner. The second network model includes a scheduling network model and an MRS cycle prediction network model. The scheduling network model is used to adjust the combination of MRS water purification equipment deployed in a timely manner according to the actual situation of the entire wastewater treatment system. The scheduling network model in the second network model is a spatiotemporal graph neural network prediction model. The spatiotemporal graph neural network prediction model is trained using second training data, which includes graphs of historical time periods, route segments with failure risks, and preset combinations of MRS water purification equipment, to obtain the trained spatiotemporal graph neural network prediction model. The MRS cycle prediction network model is used to accurately predict the recovery cycle of each MRS water purification device and transmit the data to the scheduling network model. The MRS cycle prediction network model includes two levels. The first level uses an LSTM network to learn historical operating data and outputs current and future microbial kinetic parameters that change with conditions in real time. The second level inputs the output microbial kinetic parameters into a water purification cycle calculation formula derived and transformed from the Monod equation to directly calculate the theoretical recovery cycle.

2. The method according to claim 1, wherein the formula for predicting the recovery period is: in, This indicates the predicted remaining recovery period for the MRS water purification equipment within the time period t+1. This indicates the initial contaminant concentration of the MRS water purification equipment. This indicates the critical contaminant concentration at which the water quality of the MRS water purification equipment meets the standards. It is the half-saturation constant, reflecting the affinity of microorganisms for pollutants. The predicted half-saturation constant during the time interval t+1. The dynamic maximum reaction rate, The dynamic maximum reaction rate is predicted within the time interval t+1, and η is the system efficiency correction coefficient. This is the predicted correction coefficient for the time period t+1.

3. The method according to claim 2, wherein the specific working steps of the MRS periodic prediction network model include: Data preparation and preprocessing: Collect historical operating data and corresponding recovery cycles for each MRS water purification device, and clean, align, and normalize the data; LSTM network model training; Online deployment and prediction of LSTM network models; The model is updated adaptively.

4. The method according to claim 3, wherein, The LSTM network model training includes: Historical time series data The model is then supervised to learn from the input. Utilizing the same historical period , Based on the actual observed recovery period, the equivalent recovery time during that period can be calculated in reverse. , , Parameter value; Construct a multi-layer LSTM network, using mean squared error as the loss function, and train the network to learn from... To the three mentioned , , The mapping relationship of parameter values.

5. The method according to claim 4, wherein, The online deployment and prediction of the LSTM network model includes... Deploy the trained LSTM model to the real-time system; The system collects data in real time, inputs it into the model, and obtains dynamic parameters. , , Predicted value; Dynamic parameters , , The predicted value is substituted into the cycle prediction formula to calculate and update the remaining recovery cycle of the MRS water purification equipment in real time.

6. The method according to claim 1, wherein the step of constructing a spatiotemporal map based on the total data set includes: The set of graph nodes is determined based on historical and current data associated with wastewater at multiple monitoring nodes; The graph edge set is determined based on the connection relationships of the pipeline network, the connection relationships of the sewage outlets, and the connection relationships of the inlet and outlet of the MRS water purification equipment; and the graph node set and the graph edge set are processed using a graph neural network to determine the graph spectrum.

7. The method according to claim 6, wherein the first network model is a graph neural network, and the step of using the trained first network model to process the graph and determine route segments with fault risk includes: The graph neural network is trained using first training data, which includes feature data of various blockage types in the spatiotemporal dimension. The feature data is determined based on historical and current data associated with sewage at multiple monitoring nodes. The graph is processed using a trained graph neural network to determine the blockage type, which includes slow-deposition blockage and sudden foreign body blockage. Based on the map and the congestion type, route segments with fault risk are identified.

8. A multi-source data-based MRS wastewater treatment device, used to perform the method according to any one of claims 1-7, comprising: The data acquisition module is configured to acquire a total set of data related to wastewater at multiple monitoring nodes in the target area. The total set of data includes at least a wastewater production data set, a wastewater pipeline network data set, an MRS wastewater treatment data set, and an environmental data set. The graph construction module is configured to construct a spatiotemporal graph based on the total data set. The graph includes a set of graph nodes and a set of graph edges, wherein the graph edges indicate the spatial relationship between sewage outlets, pipe networks, and the inlet and outlet of MRS water purification equipment. The risk segment identification module is configured to use a trained first network model to process the map and identify route segments with fault risk. The network scheduling module is used to adjust the combination of MRS water purification equipment put into operation in a timely manner according to the actual situation of the entire sewage treatment system. The MRS cycle prediction module is used to accurately predict the cycle of each MRS water purification device and transmit the data to the network scheduling module.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images