Distributed sewage treatment and automatic control system

Abstract

The invention discloses a distributed sewage treatment and automatic control system, and relates to the technical field of sewage automatic treatment, and the system comprises the steps: collecting the check-in reservation data and geographic coordinate information of a tourism business operation site in a target area in real time, weather rainfall forecast information, and a sewage pipe network topological structure; constructing a sewage load prediction model, and identifying check-in reservation data and meteorological rainfall forecast information of the tourism business operation site to obtain load prediction data; based on the geographic coordinate information, mapping the load prediction data of the tourism business operation site to a sewage pipe network topological structure to obtain sewage topological mapping data; determining a source position and a propagation path of load impact; constructing a treatment strategy optimization model, identifying the sewage topological mapping data and the meteorological rainfall forecast information, and dynamically generating a sewage treatment strategy; a sewage treatment strategy is transmitted to a sewage treatment actuator through a low-power-consumption wide-area internet of things communication protocol, and automatic control over sewage treatment is achieved.

Description

Technical Field

[0001] This invention relates to the field of automated wastewater treatment technology, and in particular to a decentralized wastewater treatment and automated control system. Background Technology

[0002] With the rapid development of the tourism industry and the advancement of the rural revitalization strategy, the demand for wastewater treatment facilities is increasing in scenic areas, resorts, and scattered rural towns. Traditional centralized large-scale wastewater treatment plants are unable to meet the needs of these geographically dispersed tourist destinations, which are subject to strong seasonality and significant load fluctuations. Therefore, the adoption of small-scale, decentralized wastewater treatment systems has become an inevitable trend.

[0003] However, decentralized wastewater treatment systems face key technical challenges: the wastewater discharge load of tourist venues is highly correlated with occupancy rates, activity durations, and restaurant scale. During peak periods, influent flow and pollutant concentrations can surge several times over in a short period, causing drastic changes in the system's hydraulic retention time and instantaneous overload of the biological system. This drastic and unpredictable impact is the main reason for substandard effluent quality and low operational efficiency of decentralized treatment facilities.

[0004] Furthermore, existing decentralized wastewater treatment facilities typically employ experience-based timed control or feedback control based on real-time monitoring. Feedback control is essentially a lag control; the system only begins to respond and adjust after a load shock has entered the reactor and caused changes in water quality parameters. This lag often fails to adjust in time when faced with rapid shocks, leading to excessive pollutant emissions.

[0005] Meanwhile, timed or fixed-value control cannot adapt to actual load changes, leading to excessive operation of aeration blowers and water pumps, resulting in a large amount of energy waste.

[0006] Therefore, a decentralized wastewater treatment and automated control system is proposed. Summary of the Invention

[0007] This invention provides a decentralized wastewater treatment and automated control system to solve the problems of huge fluctuations in wastewater load and the lag and inefficiency of control strategies in the prior art, thereby achieving predictive and accurate technical effects in wastewater treatment.

[0008] This invention provides a decentralized wastewater treatment and automated control system, comprising: The multi-source data acquisition module collects real-time data on occupancy reservations and geographic coordinates of tourist attractions within the target area, as well as weather and rainfall forecasts and sewage pipe network topology. The wastewater load prediction module constructs a wastewater load prediction model, identifies occupancy reservation data from tourist venues and weather and rainfall forecast information, and obtains load prediction data; the load prediction data includes wastewater flow prediction parameters and wastewater composition prediction parameters; The wastewater spatial mapping module maps the load forecast data of tourist business premises to the wastewater pipe network topology based on geographic coordinate information to obtain wastewater topology mapping data; and determines the source location and propagation path of load impacts. The wastewater treatment optimization module constructs a treatment strategy optimization model, identifies wastewater topology mapping data and meteorological rainfall forecast information, and dynamically generates wastewater treatment strategies including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands. The strategy edge execution module transmits the wastewater treatment strategy to the wastewater treatment actuator through a low-power wide-area IoT communication protocol, thereby realizing the automated control of wastewater treatment.

[0009] The check-in reservation data includes the room type, room floor, number of guests, check-in time, and check-out time. The geographic coordinate information includes latitude and longitude data and altitude data, and the meteorological and rainfall forecast information includes temperature and humidity data and rainfall data.

[0010] A wastewater load prediction model is constructed based on spatiotemporal graph convolutional networks; The wastewater load prediction model uses occupancy reservation data as dynamic node features and encodes the wastewater pipe network topology as a graph structure. Through the graph convolutional layer and gating unit inside the wastewater load prediction model, it simultaneously captures the time dependence, spatial propagation and nonlinear characteristics of wastewater load. The load prediction data obtained by the wastewater load prediction model includes wastewater flow prediction parameters and wastewater composition prediction parameters, specifically including: time-series curve prediction parameters of total influent flow, time-series prediction parameters of total chemical oxygen demand concentration, and time-series prediction parameters of total nitrogen concentration within a future preset time period.

[0011] The training process of the wastewater load prediction model adopts online incremental training and error self-correction mechanism; the specific training process includes: Offline pre-training was performed using historical operational data and historical check-in booking data; During model operation, the deviation between the predicted output value and the actual collected value is compared in real time through the built-in prediction mean square error monitoring mechanism. When the mean square error exceeds the preset tolerance threshold, the adaptive optimization process of the wastewater load prediction model is triggered. The adaptive optimization process uses the latest collected operating data to fine-tune the core weights and learning rate of the model online, so as to realize the continuous self-learning and adaptation of the model.

[0012] The wastewater spatial mapping module performs high-precision pipe network hydraulic analysis to determine the source location and propagation path of load shocks; the specific process includes: Based on the latitude and longitude data, altitude data, and sewage pipe diameter, slope, and material parameters in the sewage topology mapping data of tourist business sites, combined with historical average flow velocity, the time interval required for sewage generated from any load source to reach the treatment station is calculated. The load forecast data is timestamped according to the time interval to obtain wastewater topology mapping data, thereby spatially correlating the location of load occurrence, impact intensity and precise arrival time.

[0013] The wastewater treatment optimization module constructs a treatment strategy optimization model; the treatment strategy optimization model is constructed with ensuring that the effluent water quality meets the standards as a hard constraint and minimizing the comprehensive energy consumption of aeration blowers and water pumps as the optimization objective function.

[0014] The optimized treatment strategy model uses the predicted load arrival time and spatial source in the wastewater topology mapping data as dynamic boundary conditions. It combines the impact of meteorological and rainfall forecast information on hydraulic conditions and performs iterative calculations within a preset optimization time window to dynamically generate wastewater treatment strategies, including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands. This achieves a global optimization balance between system energy consumption and water quality stability.

[0015] The wastewater treatment actuator includes a variable frequency fan, a variable frequency water pump, and an electric regulating valve.

[0016] The beneficial effects of this invention are: 1. This invention enables source-level prediction of unstable load impacts upstream of the sewage treatment system; by predicting the precise load timing at the hotel level, it greatly improves the lead time and accuracy of the prediction; enabling subsequent optimization control to know the intensity and timing of the load before it enters the pipe network, thus providing a key decision-making basis for implementing source regulating valve commands (such as diversion, interception, or temporary storage) to cope with load impact fluctuations caused by changes in the number of hotel guests.

[0017] 2. This invention ensures the continuous high accuracy of the wastewater load prediction module during long-term operation. By using the latest actual operating performance to correct the model, it effectively corrects prediction deviations caused by changes in occupancy behavior patterns or local changes in the pipe network, greatly improving the reliability and decision-making quality of the automated control system and achieving accurate wastewater treatment prediction.

[0018] 3. The dynamically generated strategy instructions of this invention can accurately offset the impact of upcoming load shocks and rainfall runoff on the hydraulic / biological system, avoiding energy waste and water quality fluctuations caused by traditional timed or set-value control. By converting predictive information into feedforward control, the system avoids energy waste on over-aeration or ineffective hydraulic boosting, thus optimizing energy consumption. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of a decentralized wastewater treatment and automated control system according to the present invention; Figure 2 This is a schematic diagram of a decentralized wastewater treatment and automated control system according to the present invention. Figure 3 This is a data logic diagram of a decentralized wastewater treatment and automated control system according to the present invention; Figure 4 This is a data curve showing the predicted parameters and actual monitoring parameters of wastewater flow. Detailed Implementation

[0020] The above technical solutions will be described in detail below with reference to the accompanying drawings and specific embodiments to better understand them. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should be noted that, for ease of description, only the parts related to the present invention are shown in the drawings, not all of them.

[0021] Example 1: This invention proposes a decentralized wastewater treatment and automated control system, the structure of which is as follows: Figure 1 As shown, it includes: The multi-source data acquisition module collects real-time data on occupancy reservations and geographic coordinates of tourist attractions within the target area, as well as weather and rainfall forecasts and sewage pipe network topology. The wastewater load prediction module constructs a wastewater load prediction model, identifies occupancy reservation data from tourist venues and weather and rainfall forecast information, and obtains load prediction data; the load prediction data includes wastewater flow prediction parameters and wastewater composition prediction parameters; The wastewater spatial mapping module maps the load forecast data of tourist business premises to the wastewater pipe network topology based on geographic coordinate information to obtain wastewater topology mapping data; and determines the source location and propagation path of load impacts. The wastewater treatment optimization module constructs a treatment strategy optimization model, identifies wastewater topology mapping data and meteorological rainfall forecast information, and dynamically generates wastewater treatment strategies including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands. The strategy edge execution module transmits the wastewater treatment strategy to the wastewater treatment actuator through a low-power wide-area IoT communication protocol, thereby realizing the automated control of wastewater treatment.

[0022] The process of the decentralized wastewater treatment and automated control system is as follows: Figure 2 As shown.

[0023] The check-in reservation data includes the room type, room floor, number of guests, check-in time, and check-out time. The geographic coordinate information includes latitude and longitude data and altitude data, and the meteorological and rainfall forecast information includes temperature and humidity data and rainfall data.

[0024] The multi-source data acquisition module directly obtains and structures these parameters from the hotel management system, geographic information system, and meteorological service API; This invention greatly improves the accuracy and relevance of load forecasting by providing high-dimensional, multi-scale input data, especially the number of occupants and time, which directly quantifies the load intensity; the geographic information accurately correlates the spatial location of the load, laying a precise data foundation for subsequent spatial mapping and hydraulic analysis.

[0025] A wastewater load prediction model is constructed based on spatiotemporal graph convolutional networks; The wastewater load prediction model uses occupancy reservation data as dynamic node features and encodes the wastewater pipe network topology as a graph structure. Through the graph convolutional layer and gating unit inside the wastewater load prediction model, it simultaneously captures the time dependence, spatial propagation and nonlinear characteristics of wastewater load. The load prediction data obtained by the wastewater load prediction model includes wastewater flow prediction parameters and wastewater composition prediction parameters, specifically including: time-series curve prediction parameters of total influent flow, time-series prediction parameters of total chemical oxygen demand concentration, and time-series prediction parameters of total nitrogen concentration within a future preset time period.

[0026] The process of constructing and identifying wastewater load prediction models includes: Historical data learning: The model is trained offline using a large amount of historical booking data and historical actual measured sewage flow / water quality data to learn the nonlinear mapping relationship between booking characteristics and actual discharge load.

[0027] Graph structure encoding: Encode the target hotel access point (node) and the upstream pipeline topology (edge).

[0028] Real-time prediction: Real-time acquired detailed check-in booking data is aligned in the time dimension and used as the dynamic feature vector input model for each node.

[0029] Predictive output: The spatial correlation of hotel clusters is captured by graph convolutional layers, and the temporal dependence of booking / check-in / check-out behavior is captured by gating units, outputting the future load time series prediction value at the hotel level.

[0030] Spatiotemporal graph convolutional networks can simultaneously process temporal and topological data. Through deep learning, they bypass the complex modeling methods of traditional hydraulic or empirical formula-based approaches, directly extracting predictive capabilities for wastewater parameters (flow rate, COD, TN) from high-dimensional, nonlinear lifestyle data, thus achieving accurate simulation of wastewater load occurrence stages. The number of occupants and time are strongly correlated factors, while floor level and room type serve as auxiliary, refined contextual features to enhance the model's discriminative power and accuracy.

[0031] To implement the above-mentioned wastewater load prediction model, the graph structure is encoded using the following rules: an adjacency matrix is ​​constructed to represent the connection relationship of the wastewater pipe network.

[0032] Specifically, the sewage outlets of each tourist business and the confluence nodes in the pipeline network are defined as nodes in the diagram. If there is a direct pipe connection between two nodes, the two nodes are considered connected, and the connection weight from the upstream node to the downstream node is set to a non-zero value according to the water flow direction; if there is no direct pipe connection between two nodes, the connection weight is set to zero. The specific magnitude of the weight is normalized according to the reciprocal of the pipe length, that is, the shorter the pipe, the greater the association weight between the nodes.

[0033] The computational logic of the graph convolutional layer is as follows: For each target node in the graph, the model performs a feature aggregation operation. This involves weighting and summing the feature data (including occupancy count, booking duration, etc.) of all neighboring nodes connected to the target node at the previous time step according to the aforementioned connection weights; then, concatenating the sum with the target node's own feature data; finally, updating the spatial features of the target node through a nonlinear activation transformation of a fully connected layer. This process simulates the potential cumulative effect of passenger flow load on the pipeline network space.

[0034] The gated unit specifically adopts a gated linear unit structure, which contains two parallel computational branches: one branch is responsible for performing linear transformation on the input time series data to extract the main trend information; the other branch generates a gate coefficient between zero and one through the S-shaped growth curve function (Sigmoid function) to determine the importance of the information.

[0035] Finally, the corresponding elements of the outputs of the two branches are multiplied together to retain key time-dependent features (such as weekend check-in peaks) and filter out random noise interference.

[0036] This invention enables source-level prediction of unstable load impacts upstream of wastewater treatment systems. By predicting precise load timing at the hotel level, it greatly improves the lead time and accuracy of predictions. This allows subsequent optimization control to know the intensity and timing of loads before they enter the pipe network, thus providing a crucial decision-making basis for implementing source regulating valve commands (such as diversion, interception, or temporary storage). It is the cornerstone of achieving predictive control of the system.

[0037] The wastewater load prediction model is trained using online incremental training and an error self-correction mechanism, including: Offline pre-training was performed using historical operational data and historical check-in booking data; During model operation, the deviation between the predicted output value and the actual collected value is compared in real time through the built-in prediction mean square error monitoring mechanism. When the mean square error exceeds the preset tolerance threshold, the adaptive optimization process of the wastewater load prediction model is triggered. The adaptive optimization process uses the latest collected operating data to fine-tune the core weights and learning rate of the model online, so as to realize the continuous self-learning and adaptation of the model.

[0038] Because the customer demographics, guest habits, and seasonal activities of tourist venues change over time, subtle shifts occur in the mapping between booking data characteristics and actual workload. Online incremental training mechanisms can capture these shifts and adjust model parameters in a timely manner, ensuring the timeliness and accuracy of model predictions.

[0039] Built-in monitoring: During the actual operation of the model, the deviation between the predicted value output by the model and the actual value collected by the monitoring point of the sewage treatment plant or pipeline network is compared in real time through the built-in mean square error of prediction (MSE) monitoring mechanism.

[0040] Adaptive optimization trigger: When the MSE exceeds the preset tolerance threshold, such as when the daily average error of predicted flow or COD exceeds the preset threshold, the adaptive optimization process of the wastewater load prediction model is triggered.

[0041] Online fine-tuning: The adaptive optimization process uses the latest collected operational data, including the latest actual sewage monitoring data and the actual / latest occupancy booking data that generated the sewage load, to fine-tune the core weights and learning rate of the model online.

[0042] Online incremental training data flow: Data alignment: Ensure that the input data (latest check-in booking data) used for self-calibration is calibrated and aligned with the output labels (actual flow / components generated by check-in behavior) over time to account for pipeline transmission delays.

[0043] Incremental learning: When the error is large, the system uses the aligned historical data pairs from the most recent period (e.g., the previous 24 hours), i.e., the pre-ordered input and the actual monitoring output, as an incremental training batch to rapidly iterate and update the parameters of the wastewater load prediction model using a small learning rate.

[0044] This invention ensures the model maintains high accuracy over long-term operation. By using the latest actual operating performance to correct the model, it effectively corrects prediction deviations caused by changes in occupancy behavior patterns or local changes in the pipeline network, greatly improving the reliability and decision-making quality of the automated control system.

[0045] The wastewater spatial mapping module performs high-precision pipe network hydraulic analysis to determine the source location and propagation path of load shocks; the specific process includes: Based on the latitude and longitude data, altitude data, and sewage pipe diameter, slope, and material parameters in the sewage topology mapping data of tourist business sites, combined with historical average flow velocity, the time interval required for sewage generated from any load source to reach the treatment station is calculated. The load forecast data is timestamped according to the time interval to obtain wastewater topology mapping data, thereby spatially correlating the location of load occurrence, impact intensity and precise arrival time.

[0046] Operation process of wastewater spatial mapping module: Data input: Load forecast data (flow rate / composition), source latitude and longitude / altitude, and pipeline geometric parameters.

[0047] Hydraulic calculation: Using Manning's formula or a similar hydraulic model, combined with pipe diameter and slope, and referring to historical average flow velocity, calculate the flow velocity of the current pipe section.

[0048] Specifically, the common form of the Manning formula is:

[0049] in, It's speed; It is a conversion constant, with a value of 1 in the International System of Units (SI). Roughness factor is a coefficient that comprehensively reflects the impact of pipe wall roughness on water flow. Its value is generally obtained from experimental data and can be selected by referring to a table when in use. It is the hydraulic radius, the ratio of the fluid cross-sectional area to the wetted perimeter. The wetted perimeter refers to the perimeter of the fluid in contact with the pipe cross-section, excluding the perimeter in contact with air. This refers to the slope of the pipe.

[0050] Time calibration: Calculate the total travel time from the source point to the treatment station along the pipeline path.

[0051] Mapping output: The load forecast data is calibrated to achieve spatial correlation between the location of load occurrence, impact intensity and precise arrival time.

[0052] Based on hydraulic principles and flow tracking, there is a delay in the transmission of wastewater in the pipe network; through hydraulic analysis, the transmission process of wastewater under gravity flow or pressure flow can be simulated. Time calibration ensures that the data received by the treatment optimization module is when the load impact arrives at the treatment station, rather than when it occurs.

[0053] The specific steps for the high-precision pipeline hydraulic analysis and timestamp calibration are as follows: The first step is to perform segmented dynamic flow velocity calculation; The system does not use a fixed average flow velocity, but dynamically calculates the flow velocity based on the predicted flow rate. For each section of pipe in the network, based on the pipe diameter, laying slope, and roughness coefficient, it uses open channel fluid dynamics principles, such as Manning's formula, to calculate the liquid level height at the current predicted flow rate; thus, it derives the actual water velocity at that flow rate. The larger the flow rate, the higher the liquid level, and the faster the corresponding flow velocity.

[0054] The second step is to perform segment-by-segment transmission time accumulation; Starting from the sewage source of any tourist business, trace downstream along the pipeline network topology. Divide the length of each pipeline segment by the calculated actual water flow velocity for that segment to obtain the transmission time required for the sewage to flow through that segment. Add up the transmission times of all pipeline segments along the path to obtain the total lag time for the sewage to reach the treatment plant.

[0055] The third step is to perform time axis translation calibration; The wastewater load data sequence generated from source prediction is shifted backward on the time axis by the calculated total lag time. For example, if a hotel is predicted to generate peak wastewater for washing up at 8:00 AM, and the calculated total transmission time is 40 minutes, the system will mark the load surge as arriving at the wastewater treatment plant at 8:40 AM, thus achieving a precise correspondence between time and space.

[0056] This invention solves the problem of time lag in load forecast data, enabling precise timing of load forecasts. This is crucial for optimizing modules, especially influent booster pumps and source regulating valves, to implement time-based pre-control strategies, effectively preventing water quality fluctuations caused by load shocks within the treatment system.

[0057] The wastewater treatment optimization module constructs a treatment strategy optimization model; the treatment strategy optimization model is constructed with ensuring that the effluent water quality meets the standards as a hard constraint and minimizing the comprehensive energy consumption of aeration blowers and water pumps as the optimization objective function.

[0058] The optimized treatment strategy model uses the predicted load arrival time and spatial source in the wastewater topology mapping data as dynamic boundary conditions. It combines the impact of meteorological and rainfall forecast information on hydraulic conditions and performs iterative calculations within a preset optimization time window to dynamically generate wastewater treatment strategies, including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands. This achieves a global optimization balance between system energy consumption and water quality stability.

[0059] The construction details of the optimization model for the processing strategy are as follows: First, a comprehensive energy consumption calculation logic is established as the optimization objective. Comprehensive energy consumption is defined as the sum of the cumulative power consumption of all executing devices within a future prediction period. Specifically, the power consumption of the aeration blower is described as a nonlinear positive correlation function between the dissolved oxygen setpoint and the influent pollutant concentration; that is, the higher the dissolved oxygen setpoint and the greater the influent pollutant concentration, the greater the required blower power. The power consumption of the water pump is described as a direct proportional function of the product of the pumping flow rate and the lifting head. The goal of the optimization solver is to find a set of control sequences that minimizes the sum of the aforementioned cumulative power consumption.

[0060] Secondly, water quality compliance is set as a rigid constraint boundary. Specifically, this means that the chemical oxygen demand (COD) and total nitrogen (TN) concentrations in the effluent at every future moment predicted by the model must be strictly lower than the prescribed emission standard limits. In addition, physical constraints on the equipment must be set, including maximum and minimum frequency limits for blowers, maximum flow limits for water pumps, and ensuring that the dissolved oxygen concentration in the bioreactor is maintained above the minimum level required to sustain microbial activity.

[0061] The processing strategy optimization model is built on the Model Predictive Control (MPC) framework.

[0062] Core Modeling: The model includes an internal predictive model that describes the biological wastewater treatment process. This internal predictive model is built on the activated sludge model (ASM), such as ASM1 or ASM3, and is supplemented by a hydraulic model to describe the dynamic changes within the reactor and sedimentation tank.

[0063] The ASM model describes the growth, decay, pollutant degradation, nitrification, and denitrification processes of microorganisms through a set of differential equations; it provides the physical-biochemical basis for predicting effluent quality (COD, TN) and analyzing the impact of control variables (DO, reflux ratio).

[0064] Water quality constraints: During the forecast period, all effluent water quality indicators must strictly meet the discharge standards.

[0065] The objective function (minimizing energy consumption) aims to minimize the total combined energy consumption over the entire optimization time window. This includes the energy consumption of aeration blowers and water pumps, with water pump energy consumption comprising the sum of the influent lift pump and the return pump. Aeration blower energy consumption is directly related to the required DO concentration setpoint; water pump energy consumption is related to the lift / return flow rate and head. The optimization model balances water quality and energy consumption by finding the sequence of control variables that minimizes energy consumption.

[0066] The expression for the optimization objective function is:

[0067] in, This indicates minimizing energy consumption; This indicates an optimized time window; Indicates total energy consumption; This indicates the wastewater treatment strategy, including dissolved oxygen concentration setpoint, mixed liquor sludge return ratio, influent booster pump start-stop sequence, and source regulating valve commands. Total energy consumption It is the sum of the energy consumption of the aeration fan and the water pump:

[0068] in, Indicates the energy consumption of the aeration fan; This indicates the energy consumption of the water pump.

[0069] The core of this processing strategy optimization module is to achieve predictive control by utilizing feedforward information.

[0070] Dynamic boundary condition identification and input: Wastewater topology mapping data identification, treatment strategy optimization model identification, and receiving timestamp-calibrated predicted load information as dynamic input boundary conditions for the system. Predict inflow flow timing: Informs when the pump's start-stop timing needs to be adjusted.

[0071] Predicting influent water quality timing: Informing the bioreactor when it will face shock loads, guiding the pre-adjustment of DO and reflux ratio.

[0072] Meteorological rainfall forecast identification: Identify rainfall data and convert it into its impact on hydraulic conditions.

[0073] Rainfall-induced infiltration of external water leads to increased inflow and water dilution, converting rainfall into additional hydraulic load and incorporating it into the forecast of inflow time series.

[0074] Iterative computation and strategy generation: The model performs rolling iterative computation within a preset optimization time window (e.g., 4-8 hours).

[0075] Step I: Predicting the effluent quality and energy consumption trajectory under different control sequences in the future time using the ASM prediction model based on dynamic boundary conditions at the current time.

[0076] Step II: Optimization Solution. Using a solver, such as Sequential Quadratic Programming (SQP) or Genetic Algorithm (GA), the optimal control sequence that satisfies all hard constraints and minimizes the objective function is found. For example, within each computational step, the Sequential Quadratic Programming algorithm first approximates the nonlinear wastewater treatment biochemical reaction process as a local quadratic function problem. Then, within this simplified local range, it searches for control parameters that minimize energy consumption and satisfy water quality constraints, such as specific dissolved oxygen and reflux ratio values. Through multiple iterative approximations, the optimal control strategy is finally determined.

[0077] Step III: Strategy Implementation: Only the control command of the first time step in the optimal control sequence is output as the wastewater treatment strategy and passed to the actuator.

[0078] Components of dynamic policy output: Dissolved oxygen concentration setpoint: Adjust the aeration intensity in advance to cope with predicted changes in organic load.

[0079] Mixed liquor sludge return ratio: Pre-adjust sludge concentration to enhance the system's shock resistance.

[0080] Start-up and shutdown sequence of inlet booster pumps: Based on the predicted flow rate and the buffering capacity of downstream water treatment facilities, peak flow rates are smoothed or temporarily stored.

[0081] Source regulating valve command: In response to high-intensity impacts, such as peak tourist seasons or sudden large loads, the diversion or shut-off valves at the source can be remotely adjusted before the load enters the pipeline network to achieve decentralized control of the load source.

[0082] Classical Model Predictive Control (MPC) uses a system model to predict the system state over a future period at each time step, and solves for the optimal control sequence while satisfying all constraints. This establishes a criterion for optimal system balance. Water quality compliance is taken as an inviolable bottom line to ensure environmental benefits; simultaneously, energy minimization is used as an economic indicator, achieving a global optimal balance between environmental and economic benefits.

[0083] The combination of feedforward and feedback control allows the model to take proactive control measures in advance by utilizing the predicted load shocks (feedforward information) and the impact of rainfall on hydraulic conditions. For example, the model can reduce DO or adjust the reflux ratio in advance before the load arrives, rather than waiting for the water quality to deteriorate before making corrections, thereby achieving global optimization of system energy consumption and water quality stability.

[0084] This invention achieves true predictive control. The dynamically generated strategy instructions can accurately offset the upcoming load shocks and disturbances to the hydraulic / biochemical system caused by rainfall and runoff, avoiding the energy waste and water quality fluctuations caused by traditional timed or fixed-value control.

[0085] The dynamic iterative process achieves global optimization and balance. By transforming predictive information into feedforward control, the system avoids energy waste on excessive aeration or ineffective hydraulic boosting. In particular, the addition of source regulating valve commands enables the system to homogenize the load, significantly reducing the operational pressure on the wastewater treatment plant, thereby achieving the maximum reduction in system energy consumption while ensuring stable water quality.

[0086] The data logic of the decentralized wastewater treatment and automated control system is as follows: Figure 3 As shown.

[0087] The wastewater treatment actuator includes a variable frequency fan, a variable frequency water pump, and an electric regulating valve.

[0088] Furthermore, the system is equipped with a source buffer storage tank at the end of the main sewage pipe of the tourist business premises, before it connects to the municipal pipe network. The source regulating valve is specifically implemented as a shut-off valve installed on the main pipe leading to the municipal pipe network, and a diversion valve installed on the bypass pipe leading to the source buffer storage tank.

[0089] Based on the above hardware structure, the execution logic of the source regulating valve command is as follows: when the sewage load prediction module warns of an impending extreme peak load exceeding the pipeline network's transport capacity or the treatment station's processing capacity, the system issues a peak shaving and temporary storage command, controls the opening of the interceptor valve to decrease, and simultaneously opens the diversion valve to temporarily introduce some high-concentration or high-flow sewage into the source buffer storage tank for storage, thereby reducing the instantaneous impact intensity entering the pipeline network; after the predicted load peak has passed and the pipeline network load has decreased, a staggered discharge command is issued, and the temporarily stored sewage is slowly and evenly pumped into the pipeline network using the booster pump in the storage tank, achieving source homogenization regulation of the load.

[0090] Actuators are the key physical units for achieving refined and variable control. Variable frequency technology enables continuous adjustment of the power output of fans and pumps, rather than simply starting and stopping them, thus achieving precise energy control. Electric regulating valves provide remote and automated hydraulic control capabilities. These actuators offer high-precision execution capabilities, faithfully and efficiently executing complex and dynamic control strategies generated by optimization models, ensuring that the energy efficiency and effectiveness of the control strategies are realized at the physical level.

[0091] Example 2:

[0092] A decentralized wastewater treatment and automated control system, the system comprising: The multi-source data acquisition module collects real-time data on occupancy reservations and geographic coordinates of tourist venues within the target area, as well as weather and rainfall forecasts and sewage pipe network topology.

[0093] The reservation data includes the room type, room floor, number of guests, check-in time, and check-out time; the geographic coordinate information includes latitude and longitude data and altitude data; and the weather and rainfall forecast information includes temperature, humidity data, and rainfall data.

[0094] In addition to the data mentioned above, the multi-source data acquisition module in this embodiment also collects the following parameters for model training and calibration: Real-time operational data: Real-time water temperature, pH value, and oxidation-reduction potential (ORP) data at the inlet and outlet of the wastewater treatment plant.

[0095] Pipeline monitoring data: Real-time liquid level height at key manifold nodes (obtained via ultrasonic level gauges).

[0096] The wastewater load prediction module constructs a wastewater load prediction model, identifies occupancy reservation data and meteorological rainfall forecast information from tourist venues, and obtains load prediction data; the load prediction data includes wastewater flow prediction parameters and wastewater composition prediction parameters.

[0097] A wastewater load prediction model is constructed based on spatiotemporal graph convolutional networks; The wastewater load prediction model uses occupancy reservation data as dynamic node features and encodes the wastewater pipe network topology as a graph structure. Through the graph convolutional layer and gating unit inside the wastewater load prediction model, it simultaneously captures the time dependence, spatial propagation and nonlinear characteristics of wastewater load. The load prediction data obtained by the wastewater load prediction model includes wastewater flow prediction parameters and wastewater composition prediction parameters, specifically including: time-series curve prediction parameters of total influent flow, time-series prediction parameters of total chemical oxygen demand concentration, and time-series prediction parameters of total nitrogen concentration within a future preset time period.

[0098] Specifically, the wastewater load prediction model uses a three-layer spatiotemporal graph convolutional block, each containing a graph convolutional layer (GCN) and a gated linear unit (GLU), with the hidden layer dimension uniformly set to 128.

[0099] The output layer is a fully connected layer that outputs time-series predicted values ​​of total influent flow, total chemical oxygen demand (COD), and total nitrogen (TN) for the next 4 hours at 15-minute intervals (a total of 16 time points).

[0100] Dynamic node feature encoding: Number of guests and time of stay: The number of guests booked is normalized to Min-Max.

[0101] The booked check-in / check-out times are converted into a 24-dimensional vector using one-hot encoding, representing the probability weight of occurrence for each hour within the prediction period.

[0102] Room type / floor reservation: Discrete categorical data is mapped to an 8-dimensional dense vector through an embedding layer.

[0103] The above feature vectors are concatenated to form the dynamic input feature vector for each tourism business location node.

[0104] Graph structure encoding and connection weights: This embodiment uses adjacency matrix weights based on traffic contribution and transmission time correction; Connection weight definition: The connection weight from an upstream node to a downstream node is defined as the product of the average traffic share contributed by that node and the transmission time attenuation factor.

[0105] in: Let i be the historical average daily emission flow rate. This is the set of all upstream nodes that converge to node j; Let be the average transmission time of the sewage from node i to node j; This is the time decay constant, set to 0.5 by default. This ensures that the longer the transmission time, the lower the weight of the load information on the current node status, thus solving the accuracy problem of simply using the reciprocal of the pipe length as the weight.

[0106] Online incremental training and error self-correction: Data alignment correction: Considering the non-point source and multi-source mixed nature of the load, the triggering conditions and data alignment logic of the built-in mean square error of prediction (MSE) monitoring mechanism have been adjusted as follows: The system no longer attempts to precisely align a single booking record with a single emission value. Instead, it treats all booking inputs within the forecast time window as a batch and compares its predicted cumulative total load with the actual cumulative total load within the same time window after calibration with the actual lag time.

[0107] Furthermore, data alignment can be achieved using a dynamic time window translation alignment method; to construct an effective dataset for online incremental training, the following three steps are performed: Step 1: Construct the input feature batch and define a time window, such as 1 hour. The system selects the set of booking data prior to the current moment. Input the active occupancy booking vectors of all tourist venues within this time window, and real-time weather data, representing the set of source behaviors that generated sewage.

[0108] Step 2: Calculate the hydraulic lag time. Using the high-precision pipe network hydraulic analysis in the wastewater spatial mapping module, calculate the average hydraulic transmission time from the centroid of the source area to the inlet of the wastewater treatment plant, i.e., the lag time. This time is dynamically calculated based on the current actual flow rate, rather than a fixed value.

[0109] Step 3: Construct output label batches. On the timeline of the actual monitoring data, capture the starting point of the time as the current moment plus the average hydraulic transmission time, and the ending point of the time as the current moment plus the average hydraulic transmission time plus the inflow rate and water quality data of the time window.

[0110] Translation logic: The calculated lag time is shifted backward by the monitoring window.

[0111] Data aggregation: Calculate the cumulative total flow and total amount of pollutants within the shift window, representing the physical result of the input batch's behavior reaching the treatment station after being transmitted through the pipeline network.

[0112] Adaptive optimization trigger: When the daily average relative error exceeds 5% for three consecutive days, the adaptive optimization process is triggered.

[0113] Online fine-tuning: The adaptive optimization process uses the aligned data of the most recent 7 days as incremental training batches and uses a small learning rate of 0.001 to fine-tune the model online.

[0114] The wastewater spatial mapping module maps the load forecast data of tourist business premises to the wastewater pipe network topology based on geographic coordinate information to obtain wastewater topology mapping data; and determines the source location and propagation path of load impact.

[0115] The specific steps for the high-precision pipeline hydraulic analysis and timestamp calibration are as follows: The first step is to perform segmented dynamic flow velocity calculation; The system dynamically calculates the flow velocity based on the predicted flow rate. For each section of pipe in the network, based on the pipe's diameter, laying slope, and roughness coefficient, it uses open channel fluid dynamics principles, such as the Manning formula, to calculate the liquid level height at the current predicted flow rate; thus, it derives the actual water velocity at that flow rate. The higher the flow rate, the higher the liquid level, and the faster the corresponding flow velocity.

[0116] It should be noted that the segmented dynamic flow velocity calculation is based on the flow value generated by the source prediction and the flow value superimposed segment by segment from upstream to downstream, and is not based on the downstream arrival flow that has not yet been determined.

[0117] The second step is to perform segment-by-segment transmission time accumulation; Starting from the sewage source of any tourist business, trace downstream along the pipeline network topology. Divide the length of each pipeline segment by the calculated actual water flow velocity for that segment to obtain the transmission time required for the sewage to flow through that segment. Add up the transmission times of all pipeline segments along the path to obtain the total lag time for the sewage to reach the treatment plant.

[0118] The third step is to perform time axis translation calibration; The wastewater load data sequence generated from source prediction is shifted backward on the time axis by the calculated total lag time. For example, if a hotel is predicted to generate peak wastewater for washing up at 8:00 AM, and the calculated total transmission time is 40 minutes, the system will mark the load surge as arriving at the wastewater treatment plant at 8:40 AM, thus achieving a precise correspondence between time and space.

[0119] Furthermore, the segmented dynamic flow velocity calculation not only relies on Manning's formula and predicted flow rate but also incorporates real-time liquid level monitoring data. For pipe segments with liquid level monitoring, such as at important confluence nodes, the flow velocity calculation prioritizes the actual water depth calculated from the measured liquid level, combined with Manning's formula to calculate the current actual flow velocity. For pipe segments without monitoring, the predicted flow rate is used in conjunction with pipe diameter, slope, and roughness coefficient (0.013) to calculate the water depth and flow velocity. Real-time transmission time calculation: During the accumulation of segmented transmission time, the above-mentioned corrected flow velocity is used in real-time for calculation.

[0120] The final mapping result obtained from the wastewater topology mapping data should include the following three key pieces of information to ensure accurate timing of the load: Source location: The unique ID and geographic coordinates (latitude, longitude / altitude) of the tourist business location.

[0121] Impact intensity: time series curves for predicted flow rate, COD, and TN.

[0122] Precise arrival time: The timestamp sequence obtained by shifting the source prediction time series backward by the total lag time and then calibrating it.

[0123] The wastewater treatment optimization module constructs a treatment strategy optimization model, identifies wastewater topology mapping data and meteorological rainfall forecast information, and dynamically generates wastewater treatment strategies including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands.

[0124] The optimized treatment strategy model uses the predicted load arrival time and spatial source in the wastewater topology mapping data as dynamic boundary conditions. It combines the impact of meteorological and rainfall forecast information on hydraulic conditions and performs iterative calculations within a preset optimization time window to dynamically generate wastewater treatment strategies, including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent lift pump start-stop sequences, and source regulating valve commands. This achieves a global optimization balance between system energy consumption and water quality stability.

[0125] The strategy edge execution module transmits the wastewater treatment strategy to the wastewater treatment actuator through a low-power wide-area IoT communication protocol, thereby realizing the automated control of wastewater treatment.

[0126] The policy edge execution module in this embodiment defines the communication protocol and control interface: Communication Protocol: Policy commands are transmitted from the cloud optimization server to the edge execution controller of the distributed processing station via the NB-IoT protocol.

[0127] Strategy parsing and execution interface: After receiving the policy, the edge controller parses the instruction packet using the MQTT protocol.

[0128] The parsed control commands are sent to the lower-level actuators via the industry-standard Modbus / TCP protocol.

[0129] The dissolved oxygen setpoint and reflux ratio command control the frequency of the variable frequency fan and variable frequency reflux pump through the analog output interface.

[0130] The start-stop sequence of the inlet booster pump and the commands to the source regulating valve are controlled by the contactor and the electric regulating valve through the digital output interface.

[0131] To verify the control effect of the control system of the present invention, comparative experiments were conducted, including: Control group 1: adopts traditional timed control; the aeration pressure and booster pumps operate according to a preset schedule, such as fully open during the day and half open at night, without adjustment based on real-time water quality; Control Group 2: Conventional PID control; relying on a DO sensor installed inside the biological treatment tank and a level gauge at the inlet for feedback control. When the sensor detects that the DO concentration is lower than the set value, the control frequency is increased through the PID algorithm.

[0132] Experimental group: The automated control system described in this invention.

[0133] Background: During the peak tourist season, 8:30-9:30 AM is the peak time for guests to wash up and check out, resulting in high traffic and high concentration of visitors.

[0134] Forecast time window: 08:00-12:00 (next 4 hours); Time step: 15 minutes (16 points in total); Hydraulic lag: According to hydraulic analysis and calculation of the pipeline network, it takes about 45 minutes (3 time steps) from the source to the treatment station.

[0135] In the experimental group, the automated control system described in this invention employs a wastewater load prediction module to identify occupancy reservation data of tourist venues, weather and rainfall forecast information, and the topology of the wastewater pipe network to obtain load prediction data. The data for wastewater flow prediction parameters and actual monitoring parameters are shown in Table 1; the data curves for wastewater flow prediction parameters and actual monitoring parameters are shown in... Figure 4 As shown.

[0136] Table 1 Comparison of predicted and actual wastewater flow parameters

[0137] Data was collected and processed by simulating the water usage of tourist venues under real-world conditions, and the results are shown in Table 2.

[0138] Table 2. Statistical table of effluent water quality stability under different control strategies.

[0139] The overall compliance rate is the ratio of the number of data points that meet the requirements to the total number of data points.

[0140] Subsequently, the wastewater treatment optimization module identifies wastewater topology mapping data and meteorological rainfall forecast information, and dynamically generates wastewater treatment strategies including dissolved oxygen concentration setpoints, mixed liquor sludge return ratios, influent booster pump start-stop sequences, and source regulating valve commands.

[0141] The experimental group not only achieved a comprehensive compliance rate of 99.9%, but also demonstrated greater stability (standard deviation) of the effluent quality compared to the control group, thus verifying the effectiveness of the invention in ensuring effluent quality compliance as a rigid constraint.

[0142] To realize the above-mentioned decentralized wastewater treatment and automated control logic, this embodiment constructs a hierarchical Internet of Things (IoT) hardware architecture, which mainly includes a sensing layer, an edge control layer, and an execution layer.

[0143] The specific hardware configuration and connection scheme are as follows: System hardware topology: The system is physically deployed with two core nodes: a source monitoring node, deployed at the sewage outlets of tourist business premises, and a central treatment station node, deployed at the site of decentralized sewage treatment facilities. The source monitoring node is responsible for collecting source flow data and executing diversion commands; the central treatment station node is responsible for receiving cloud-based policies and executing core process controls.

[0144] Cloud server: Deploy the above-mentioned wastewater load prediction model and treatment strategy optimization model.

[0145] To meet the requirements of low-power wide-area IoT and frequency conversion control mentioned in this invention, an edge computing and communication gateway is selected, integrating an NB-IoT / 4G full network module to receive policy instructions (DO setpoint, return ratio, etc.) sent from the cloud; it provides two RS-485 interfaces (Modbus RTU protocol) and one Ethernet interface (Modbus TCP protocol) for connecting PLC controllers and online instruments.

[0146] It should be understood that the embodiments disclosed in this invention and the above description enable those skilled in the art to implement this invention. However, this invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention, and should all be included within the protection scope of this invention.

Claims

1. A decentralized wastewater treatment and automation control system, characterized in that, The system comprises: A multi-source data acquisition module acquires in real time the check-in reservation data and geographic coordinate information of the tourist operating place in the target area, meteorological rainfall forecast information, and sewage pipe network topology; the check-in reservation data includes the reservation room type, reservation room floor, reservation number of occupants, reservation check-in time, and reservation check-out time; A sewage load prediction module constructs a sewage load prediction model, identifies the check-in reservation data, meteorological rainfall forecast information, and sewage pipe network topology of the tourist operating place, and obtains load prediction data; the load prediction data includes sewage flow prediction parameters and sewage composition prediction parameters; A sewage space mapping module maps the load prediction data of the tourist operating place to the sewage pipe network topology based on the geographic coordinate information, and obtains sewage topology mapping data; the source location and propagation path of the load impact are determined; A sewage treatment optimization module constructs a treatment strategy optimization model, identifies the sewage topology mapping data and meteorological rainfall forecast information, and dynamically generates a sewage treatment strategy including the dissolved oxygen concentration set value, mixed liquid sludge reflux ratio, water inlet lifting pump start-stop timing, and source adjustment valve instruction; A strategy edge execution module transmits the sewage treatment strategy to the sewage treatment executor through a low-power wide-area Internet of Things communication protocol, and realizes the automatic control of sewage treatment.

2. The distributed wastewater treatment and automated control system of claim 1, wherein: The geographic coordinate information includes longitude and latitude data and elevation data, and the meteorological rainfall forecast information includes temperature and humidity data and rainfall data.

3. A decentralized wastewater treatment and automated control system as claimed in claim 2, wherein, It comprises: According to the spatio-temporal graph convolution network, a sewage load prediction model is constructed; The sewage load prediction model encodes the check-in reservation data as dynamic node features and the sewage pipe network topology as a graph structure; Through the graph convolution layer and the gate unit inside the sewage load prediction model, the time dependence, spatial propagation, and non-linear characteristics of the sewage load are captured simultaneously; Then, the data is corrected according to the meteorological rainfall forecast information; The load prediction data obtained by the sewage load prediction model includes sewage flow prediction parameters and sewage composition prediction parameters, specifically including: time series curve prediction parameters of total inflow, concentration time series prediction parameters of total chemical oxygen demand, and concentration time series prediction parameters of total nitrogen in a future preset time period.

4. A decentralized wastewater treatment and automated control system as claimed in claim 3, wherein, The training process of the sewage load prediction model adopts an online incremental training and error self-correction mechanism; The specific training process includes: Using historical operation data and historical check-in reservation data for offline pre-training; During the model running process, the built-in prediction mean square error monitoring mechanism is used to compare the deviation between the predicted output value and the actual collected value in real time; When the mean square error exceeds the preset tolerance threshold, the adaptive optimization process of the sewage load prediction model is triggered; the latest collected operation data is used to fine-tune the core weight and learning rate of the model, realizing the continuous self-learning and adaptation of the model.

5. A decentralized wastewater treatment and automated control system as claimed in claim 2, wherein, It comprises: The sewage space mapping module performs pipe network hydraulic analysis to determine the source location and propagation path of the load impact; The specific process includes: Based on the longitude and latitude data of the tourism operating place, the altitude data, and the sewage topology mapping data including the sewage pipe network diameter, slope, and material parameters; combined with the historical average flow rate, the time interval required for the sewage generated by any load source point to reach the treatment station is calculated; The load prediction data is time-stamped calibrated according to the time interval, and the sewage topology mapping data is obtained, so as to spatially correlate the occurrence position, impact strength, and accurate arrival time of the load.

6. A decentralized wastewater treatment and automated control system as claimed in claim 1, wherein, Comprise: The sewage treatment optimization module constructs a treatment strategy optimization model; the treatment strategy optimization model is constructed with the guarantee of the effluent water quality reaching the standard as a hard constraint condition and with the minimization of the comprehensive energy consumption of the aeration fan and the water pump as an optimization objective function.

7. A decentralized wastewater treatment and automated control system as claimed in claim 6, wherein, Comprise: The treatment strategy optimization model takes the predicted load arrival time and spatial source in the sewage topology mapping data as a dynamic boundary condition, combines the influence of the meteorological rainfall forecast information on the hydraulic condition, and iteratively calculates within a preset optimization time window to dynamically generate the sewage treatment strategy including the dissolved oxygen concentration set value, the mixed liquid sludge return ratio, the water inlet lifting water pump start-stop time sequence, and the source regulating valve instruction, so as to realize the global optimization balance of the system energy consumption and the water quality stability.

8. A decentralized wastewater treatment and automated control system as claimed in claim 7, wherein, Comprise: The sewage treatment executor comprises a variable-frequency fan, a variable-frequency water pump, and an electric regulating valve.

Images